{
  "name": "Which Claude model for a resume: one resume, one posting, four models, November 2026",
  "publisher": "JobShifu",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "temporalCoverage": "2026-11-13",
  "post": "https://jobshifu.com/blog/claude-resume-tailoring",
  "method": "A fictional resume (Alex Morgan, senior backend engineer) and one live posting (Robinhood, Senior Software Engineer, Backend, https://boards.greenhouse.io/robinhood/jobs/7264931, fetched 2026-11-13) sent to claude-fable-5-2, claude-opus-5-5, claude-sonnet-5-5 and claude-haiku-5-5 through the Messages API (Batch), no system prompt, each model's default effort, one run each. Two prompts: a bare \"Tailor my resume to this job posting\" and the hard-rules Prompt 6 from https://jobshifu.com/blog/tailor-resume-to-job-description-with-ai-free. Each tailored resume was read by JobShifu's Calibration Engine code: figureSupport (numbers stated nowhere in the resume), the jd-echo scanner (posting terms the resume never writes) and the production drift judge (a line claiming more than the resume lines it draws on). Lines were matched to resume lines by shared content words. Every drift flag was then read by hand against the resume; driftFlagsThatHold counts the ones that stand.",
  "notTested": "One resume, one posting, one run per cell. A close-fit senior backend resume; a career changer against a stretch posting was not tested. Claude.ai chat applies its own system prompt and settings, which the API run does not.",
  "results": [
    {
      "model": "claude-fable-5-2",
      "prompt": "bare",
      "resumeLines": 33,
      "keptWordForWord": 22,
      "inventedNumbers": 0,
      "postingEchoLines": 0,
      "driftFlags": 1,
      "driftFlagsThatHold": 1,
      "notWrittenItems": null,
      "outputTokens": 3791,
      "batchCostUsd": 0.0871,
      "flaggedLines": [
        {
          "text": "Senior backend engineer with 9 years building and operating distributed systems in production, primarily in Python and Go. Experience spans low-latency services, streaming billing pipelines, and REST APIs, with a focus on correctness, reliability, and reducing operational risk. Leads multi-quarter technical projects and mentors engineers.",
          "judge": "'distributed systems' is broader than 'backend services and developer platforms'; 'low-latency services' and 'streaming billing pipelines' are supported by sources 2-4; 'REST APIs' supported by source 4; 'Leads multi-quarter technical projects' adds a new claim not stated in the profile summary; 'mentors engineers' adds a new claim not in sources."
        }
      ],
      "tailoredResume": "# Alex Morgan\n\nSan Francisco, CA · alex.morgan.test@example.com · (415) 555-0142\nlinkedin.com/in/alexmorgantest · github.com/alexmorgantest · Authorized to work in the US\n\n## Summary\n\nSenior backend engineer with 9 years building and operating distributed systems in production, primarily in Python and Go. Experience spans low-latency services, streaming billing pipelines, and REST APIs, with a focus on correctness, reliability, and reducing operational risk. Leads multi-quarter technical projects and mentors engineers.\n\n## Experience\n\n### Senior Software Engineer — Northwind Data (2022 - Present)\n\n- Led the migration of the billing pipeline from a nightly batch job to a streaming Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.\n- Cut p95 API latency from 820ms to 210ms in the orders service by adding a read-through Redis cache and removing three N+1 query patterns.\n- Introduced service-level objectives for the six highest-traffic endpoints and wired the error budget into the release checklist, cutting rollbacks from 9 to 2 a quarter.\n- Designed and shipped the multi-region failover runbook and led two live game-day exercises with the on-call rotation.\n- Ran the vendor evaluation for the observability stack and led the migration of 40 services onto OpenTelemetry tracing over two quarters.\n- Mentored 4 engineers, two of whom were promoted to senior within 18 months.\n- Rewrote the customer data export as a resumable job, taking the failure rate on exports over 2GB from roughly 30% to under 1%.\n\n### Software Engineer — Cobalt Systems (2019 - 2022)\n\n- Built an internal feature-flag service in Go, adopted by 14 teams and serving approximately 40,000 evaluations per second at peak.\n- Replaced a hand-rolled job scheduler with Temporal, eliminating the duplicate-charge class of bugs and retiring about 4,000 lines of bespoke retry logic.\n- Owned the PostgreSQL upgrade from 11 to 14 across six production clusters with zero customer-visible downtime.\n- Reduced CI pipeline duration from 26 minutes to 9 minutes by parallelizing the integration test suite and caching container layers.\n- Led the platform team's on-call rotation and wrote 11 of the team's runbooks.\n- Built the data retention tooling that brought the analytics warehouse into GDPR compliance ahead of the audit deadline.\n\n### Junior Software Engineer — Trellis Apps (2017 - 2019)\n\n- Implemented the REST API for the scheduling product used by roughly 300 clinics.\n- Wrote the company's first integration test harness, raising coverage from 12% to 68%.\n- Built the appointment reminder service on Twilio, cutting no-show rates by 22% across the pilot group of 40 clinics.\n\n### Software Engineering Intern — Puget Analytics (Summer 2016)\n\n- Built a Flask and D3 sales dashboard and the ETL job that backfilled three years of historical sales records.\n\n## Projects\n\n### pgslice-lite — open source (2021 - Present)\n\n- Postgres partition-management CLI in Go; 900+ GitHub stars, 23 outside contributors.\n- Wrote the migration planner that computes a zero-downtime cutover path for tables above 500GB.\n\n### Ledgerbook — side project (2020)\n\n- Double-entry bookkeeping API in Python and FastAPI, used by two small businesses.\n\n## Technical Skills\n\n- **Languages:** Python, Go, TypeScript, Rust, SQL\n- **Distributed systems and data:** Kafka, Temporal, PostgreSQL, Redis, gRPC, REST API design, Airflow, dbt, Snowflake\n- **Infrastructure:** AWS, GCP, Kubernetes, Docker, Terraform, GitHub Actions\n- **Reliability and observability:** OpenTelemetry, Grafana, SLOs and error budgets, on-call and incident runbooks\n- **Frameworks and testing:** FastAPI, Flask, React, pytest\n\n## Talks and Writing\n\n- \"Paying Down Pipeline Debt\", PyCascades 2023 (conference talk).\n- Author of the Northwind engineering blog post on multi-region failover testing, approximately 30,000 reads.\n\n## Education\n\n**B.S. Computer Science**, University of Washington, 2017\n- Minor in Statistics. Senior capstone: distributed rate limiter in Rust.\n- Teaching assistant for the undergraduate databases course, two semesters.\n\n## Certifications\n\n- AWS Certified Solutions Architect — Associate (2021, renewed 2024)\n- Certified Kubernetes Administrator (CKA), 2022\n\n## Languages\n\nEnglish (native), Spanish (professional working proficiency)"
    },
    {
      "model": "claude-fable-5-2",
      "prompt": "rules",
      "resumeLines": 21,
      "keptWordForWord": 17,
      "inventedNumbers": 0,
      "postingEchoLines": 0,
      "driftFlags": 1,
      "driftFlagsThatHold": 0,
      "notWrittenItems": 13,
      "outputTokens": 12601,
      "batchCostUsd": 0.2654,
      "flaggedLines": [
        {
          "text": "Senior software engineer with 9 years building and operating backend systems and developer platforms, with a focus on reliability. Works across Python, Go and TypeScript. Earlier work includes a feature-flag service in Go serving approximately 40,000 evaluations per second at peak (built at Cobalt Systems) and the REST API for a scheduling product used by roughly 300 clinics (implemented at Trellis Apps).",
          "judge": "The line omits 'internal' and 'adopted by 14 teams' (narrowing, faithful). However, it attributes the REST API for a scheduling product to Trellis Apps, but no source mentions Trellis Apps, a scheduling product, clinics, or a REST API — this is an added fact not present in any source."
        }
      ],
      "tailoredResume": "ALEX MORGAN\nSan Francisco, CA · alex.morgan.test@example.com · (415) 555-0142\nlinkedin.com/in/alexmorgantest · github.com/alexmorgantest · Authorized to work in the US\n\nSenior Software Engineer with backend systems and reliability experience\n\nSUMMARY\nSenior software engineer with 9 years building and operating backend systems and\ndeveloper platforms, with a focus on reliability. Works across Python, Go and\nTypeScript. Earlier work includes a feature-flag service in Go serving approximately\n40,000 evaluations per second at peak (built at Cobalt Systems) and the REST API for\na scheduling product used by roughly 300 clinics (implemented at Trellis Apps).\n\nEXPERIENCE\n\nSenior Software Engineer — Northwind Data (2022 - Present)\n- Led the migration of the billing pipeline from a nightly batch job to a streaming\n  Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.\n- Cut p95 API latency from 820ms to 210ms by adding a read-through Redis cache and\n  removing three N+1 query patterns in the orders service.\n- Mentored 4 engineers, two of whom were promoted to senior within 18 months.\n- Introduced service-level objectives for the six highest-traffic endpoints and wired\n  the error budget into the release checklist, cutting rollbacks from 9 to 2 a quarter.\n- Designed and shipped the multi-region failover runbook; led two live game-day\n  exercises with the on-call rotation.\n- Ran the vendor evaluation for the observability stack and led the migration of 40\n  services onto OpenTelemetry tracing over two quarters.\n\nSoftware Engineer — Cobalt Systems (2019 - 2022)\n- Built an internal feature-flag service in Go, adopted by 14 teams and serving\n  approximately 40,000 evaluations per second at peak.\n- Owned the PostgreSQL upgrade from 11 to 14 across six production clusters with\n  zero customer-visible downtime.\n- Replaced a hand-rolled job scheduler with Temporal, retiring about 4,000 lines of\n  bespoke retry logic and eliminating the duplicate-charge class of bugs.\n- Reduced CI pipeline duration from 26 minutes to 9 minutes by parallelizing the\n  integration test suite and caching container layers.\n- On-call rotation lead for the platform team; wrote 11 of the team's runbooks.\n\nJunior Software Engineer — Trellis Apps (2017 - 2019)\n- Implemented the REST API for the scheduling product used by roughly 300 clinics.\n- Wrote the company's first integration test harness, raising coverage from 12% to 68%.\n\nPROJECTS\n\npgslice-lite — open source (2021 - Present)\n- Postgres partition-management CLI in Go; 900+ GitHub stars, 23 outside contributors.\n- Wrote the migration planner that computes a zero-downtime cutover path for tables\n  above 500GB.\n\nLedgerbook — side project (2020)\n- Double-entry bookkeeping API in Python and FastAPI, used by two small businesses.\n\nEDUCATION\nB.S. Computer Science, University of Washington, 2017\n- Minor in Statistics. Senior capstone: distributed rate limiter in Rust.\n\nCERTIFICATIONS\nAWS Certified Solutions Architect — Associate (2021, renewed 2024) · Certified\nKubernetes Administrator (CKA), 2022\n\nTECHNICAL SKILLS\nGo, Python, TypeScript, Rust, SQL, PostgreSQL, Redis, Kafka, Temporal, gRPC, Docker,\nKubernetes, Terraform, AWS, GCP, FastAPI, pytest, GitHub Actions, OpenTelemetry, Grafana"
    },
    {
      "model": "claude-opus-5-5",
      "prompt": "bare",
      "resumeLines": 32,
      "keptWordForWord": 16,
      "inventedNumbers": 0,
      "postingEchoLines": 0,
      "driftFlags": 3,
      "driftFlagsThatHold": 3,
      "notWrittenItems": null,
      "outputTokens": 2887,
      "batchCostUsd": 0.0345,
      "flaggedLines": [
        {
          "text": "Senior backend engineer with 9 years designing, building and operating distributed systems where correctness and availability are non-negotiable, including streaming billing pipelines, payment-safe workflow orchestration, and high-concurrency services handling 40,000 requests per second. Led latency, reliability and data-integrity projects from design through rollout, and mentored engineers into senior roles. Works primarily in Python and Go, with deep PostgreSQL, Kafka and Redis experience.",
          "judge": "'payment-safe workflow orchestration' adds a domain/framing not in sources; 'high-concurrency services handling 40,000 requests per second' conflates the feature-flag service stat with a broader claim about services generally; 'correctness and availability are non-negotiable' and 'streaming billing pipelines' are reasonable restatements, but 'payment-safe workflow orchestration' is an added characterization not supported by sources at this summary level."
        },
        {
          "text": "Led the redesign of the billing pipeline from a nightly batch job to a streaming Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.",
          "judge": "Source says 'migration'; line says 'redesign'. 'Redesign' implies a broader architectural ownership than 'migration', making the claim stronger/different."
        },
        {
          "text": "Mentored 4 engineers through code reviews, design discussions and hands-on guidance. Two were promoted to senior within 18 months.",
          "judge": "Line adds 'through code reviews, design discussions and hands-on guidance' — specific activities not stated in sources."
        }
      ],
      "tailoredResume": "# Alex Morgan\n\nSan Francisco, CA · alex.morgan.test@example.com · (415) 555-0142\nlinkedin.com/in/alexmorgantest · github.com/alexmorgantest · Authorized to work in the US\n\n## Summary\n\nSenior backend engineer with 9 years designing, building and operating distributed systems where correctness and availability are non-negotiable, including streaming billing pipelines, payment-safe workflow orchestration, and high-concurrency services handling 40,000 requests per second. Led latency, reliability and data-integrity projects from design through rollout, and mentored engineers into senior roles. Works primarily in Python and Go, with deep PostgreSQL, Kafka and Redis experience.\n\n## Experience\n\n### Senior Software Engineer, Northwind Data (2022 - Present)\n\n- Led the redesign of the billing pipeline from a nightly batch job to a streaming Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.\n- Cut p95 API latency in the orders service from 820ms to 210ms by adding a read-through Redis cache and eliminating three N+1 query patterns.\n- Introduced service-level objectives for the six highest-traffic endpoints and wired the error budget into the release checklist, reducing rollbacks from 9 to 2 per quarter.\n- Rewrote the customer data export as a resumable job, reducing the failure rate on exports over 2GB from roughly 30% to under 1%.\n- Designed and shipped the multi-region failover runbook and led two live game-day exercises with the on-call rotation.\n- Ran the observability vendor evaluation and led the migration of 40 services onto OpenTelemetry tracing over two quarters.\n- Mentored 4 engineers through code reviews, design discussions and hands-on guidance. Two were promoted to senior within 18 months.\n\n### Software Engineer, Cobalt Systems (2019 - 2022)\n\n- Built an internal feature-flag service in Go serving approximately 40,000 evaluations per second at peak, adopted by 14 teams.\n- Replaced a hand-rolled job scheduler with Temporal, retiring about 4,000 lines of bespoke retry logic and eliminating the duplicate-charge class of bugs.\n- Owned the PostgreSQL upgrade from 11 to 14 across six production clusters with zero customer-visible downtime.\n- Reduced CI pipeline duration from 26 minutes to 9 minutes by parallelizing the integration test suite and caching container layers.\n- Built the data retention tooling that brought the analytics warehouse into GDPR compliance ahead of the audit deadline.\n- Led the platform team's on-call rotation and wrote 11 of the team's runbooks.\n\n### Junior Software Engineer, Trellis Apps (2017 - 2019)\n\n- Implemented the REST API for the scheduling product used by roughly 300 clinics.\n- Wrote the company's first integration test harness, raising coverage from 12% to 68%.\n- Built the appointment reminder service on Twilio, cutting no-show rates by 22% across a pilot group of 40 clinics.\n- Migrated the frontend from jQuery to React over three months alongside feature work.\n\n## Projects\n\n### pgslice-lite (open source, 2021 - Present)\n\n- Postgres partition-management CLI in Go with 900+ GitHub stars and 23 outside contributors.\n- Wrote the migration planner that computes a zero-downtime cutover path for tables above 500GB.\n\n### Ledgerbook (side project, 2020)\n\n- Double-entry bookkeeping API in Python and FastAPI, used by two small businesses.\n\n## Education\n\n**B.S. Computer Science**, University of Washington, 2017\n- Minor in Statistics. Senior capstone: distributed rate limiter in Rust.\n- Teaching assistant for the undergraduate databases course, two semesters.\n\n## Certifications\n\n- AWS Certified Solutions Architect, Associate (2021, renewed 2024)\n- Certified Kubernetes Administrator (CKA), 2022\n\n## Talks and Writing\n\n- \"Paying Down Pipeline Debt\", PyCascades 2023 (conference talk).\n- Author of the Northwind engineering blog post on multi-region failover testing, approximately 30,000 reads.\n\n## Technical Skills\n\n**Languages:** Python, Go, SQL, TypeScript, Rust\n**Data and messaging:** PostgreSQL, Redis, Kafka, Temporal, Snowflake, Airflow, dbt\n**APIs and frameworks:** REST, gRPC, FastAPI, Flask, React\n**Infrastructure and reliability:** AWS, GCP, Kubernetes, Docker, Terraform, OpenTelemetry, Grafana, GitHub Actions, pytest\n\n## Languages\n\nEnglish (native), Spanish (professional working proficiency)"
    },
    {
      "model": "claude-opus-5-5",
      "prompt": "rules",
      "resumeLines": 21,
      "keptWordForWord": 14,
      "inventedNumbers": 0,
      "postingEchoLines": 0,
      "driftFlags": 3,
      "driftFlagsThatHold": 2,
      "notWrittenItems": 10,
      "outputTokens": 9065,
      "batchCostUsd": 0.0973,
      "flaggedLines": [
        {
          "text": "Replaced a hand-rolled job scheduler with Temporal, retiring about 4,000 lines of bespoke retry logic and eliminating the duplicate-charge class of correctness bugs.",
          "judge": "Line adds 'correctness' to 'duplicate-charge class of bugs' making it 'correctness bugs'; sources say only 'duplicate-charge class of bugs' with no qualifier 'correctness'. Minor addition but the sources do not state this characterization."
        },
        {
          "text": "Improved developer productivity by reducing CI pipeline duration from 26 minutes to 9 minutes, parallelizing the integration test suite and caching container layers.",
          "judge": "Line adds 'Improved developer productivity' as an outcome claim not stated in sources; sources only state the CI duration reduction and methods. This adds a broader outcome framing not in sources."
        },
        {
          "text": "Python, Go, SQL, PostgreSQL, Redis, Kafka, Temporal, REST APIs, gRPC, FastAPI, Docker, Kubernetes, Terraform, AWS, GCP, OpenTelemetry, Grafana, GitHub Actions, pytest",
          "judge": "Line includes 'REST APIs' which does not appear in the vault skills list in source [13] (which lists grpc but not 'REST APIs' as a named skill). However REST API work is evidenced in sources. All other tools are in the vault. 'REST APIs' as a skill label is not in the vault list; flagging as an addition not in sources."
        }
      ],
      "tailoredResume": "ALEX MORGAN\nSan Francisco, CA · alex.morgan.test@example.com · (415) 555-0142\nlinkedin.com/in/alexmorgantest · github.com/alexmorgantest · Authorized to work in the US\n\nSenior Software Engineer with backend services and reliability experience\n\nSUMMARY\nSenior software engineer with 9 years building and operating backend services and\ndeveloper platforms in Python and Go, with a focus on reliability and data pipelines.\nImplemented a REST API for a scheduling product used by roughly 300 clinics.\n\nEXPERIENCE\n\nSenior Software Engineer — Northwind Data (2022 - Present)\n- Led the migration of the billing pipeline from a nightly batch job to a streaming\n  Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.\n- Cut p95 API latency from 820ms to 210ms by adding a read-through Redis cache and\n  removing three N+1 query patterns in the orders service.\n- Ran the vendor evaluation for the observability stack and led the migration of 40\n  services onto OpenTelemetry tracing over two quarters.\n- Introduced service-level objectives for the six highest-traffic endpoints and wired\n  the error budget into the release checklist, cutting rollbacks from 9 to 2 a quarter.\n- Mentored 4 engineers, two of whom were promoted to senior within 18 months.\n- Designed and shipped the multi-region failover runbook; led two live game-day\n  exercises with the on-call rotation.\n- Rewrote the customer data export as a resumable job, taking the failure rate on\n  exports over 2GB from roughly 30% to under 1%.\n\nSoftware Engineer — Cobalt Systems (2019 - 2022)\n- Built an internal feature-flag service in Go, adopted by 14 teams and serving\n  approximately 40,000 evaluations per second at peak.\n- Replaced a hand-rolled job scheduler with Temporal, retiring about 4,000 lines of\n  bespoke retry logic and eliminating the duplicate-charge class of correctness bugs.\n- Owned the PostgreSQL upgrade from 11 to 14 across six production clusters with\n  zero customer-visible downtime.\n- Improved developer productivity by reducing CI pipeline duration from 26 minutes to\n  9 minutes, parallelizing the integration test suite and caching container layers.\n- On-call rotation lead for the platform team; wrote 11 of the team's runbooks.\n\nJunior Software Engineer — Trellis Apps (2017 - 2019)\n- Implemented the REST API for the scheduling product used by roughly 300 clinics.\n- Wrote the company's first integration test harness, raising coverage from 12% to 68%.\n\nPROJECTS\n- pgslice-lite (open source, 2021 - Present): Postgres partition-management CLI in Go;\n  900+ GitHub stars, 23 outside contributors. Wrote the migration planner that computes\n  a zero-downtime cutover path for tables above 500GB.\n- Ledgerbook (side project, 2020): Double-entry bookkeeping API in Python and FastAPI,\n  used by two small businesses.\n\nTALKS AND WRITING\n- \"Paying Down Pipeline Debt\", PyCascades 2023 (conference talk).\n- Author of the Northwind engineering blog post on multi-region failover testing,\n  approximately 30,000 reads.\n\nEDUCATION\nB.S. Computer Science, University of Washington, 2017 · Minor in Statistics\n\nCERTIFICATIONS\nAWS Certified Solutions Architect — Associate (2021, renewed 2024) ·\nCertified Kubernetes Administrator (CKA), 2022\n\nTECHNICAL SKILLS\nPython, Go, SQL, PostgreSQL, Redis, Kafka, Temporal, REST APIs, gRPC, FastAPI, Docker,\nKubernetes, Terraform, AWS, GCP, OpenTelemetry, Grafana, GitHub Actions, pytest"
    },
    {
      "model": "claude-sonnet-5-5",
      "prompt": "bare",
      "resumeLines": 35,
      "keptWordForWord": 28,
      "inventedNumbers": 0,
      "postingEchoLines": 0,
      "driftFlags": 1,
      "driftFlagsThatHold": 1,
      "notWrittenItems": null,
      "outputTokens": 2784,
      "batchCostUsd": 0.0167,
      "flaggedLines": [
        {
          "text": "Senior backend engineer with 9 years building and operating production services where correctness and availability matter, including billing, payments-adjacent and data-retention systems. Experienced in low-latency, event-driven and multi-region architectures in Go, Python and TypeScript. Has led large cross-team projects from design through rollout, defined SLOs and error budgets, and mentored engineers to senior.",
          "judge": "'payments-adjacent' and 'multi-region architectures' and 'low-latency' and 'event-driven' — billing pipeline source supports event-driven and low-latency inference; 'payments-adjacent' is added framing not in sources; 'multi-region' is never mentioned in any source. These add facts/domains not stated."
        }
      ],
      "tailoredResume": "# Alex Morgan\n\nSan Francisco, CA · alex.morgan.test@example.com · (415) 555-0142\nlinkedin.com/in/alexmorgantest · github.com/alexmorgantest · Authorized to work in the US\n\n## Summary\n\nSenior backend engineer with 9 years building and operating production services where correctness and availability matter, including billing, payments-adjacent and data-retention systems. Experienced in low-latency, event-driven and multi-region architectures in Go, Python and TypeScript. Has led large cross-team projects from design through rollout, defined SLOs and error budgets, and mentored engineers to senior.\n\n## Experience\n\n### Senior Software Engineer — Northwind Data (2022 - Present)\n\n- Led the migration of the billing pipeline from a nightly batch job to a streaming Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.\n- Cut p95 API latency from 820ms to 210ms by adding a read-through Redis cache and removing three N+1 query patterns in the orders service.\n- Designed and shipped the multi-region failover runbook; led two live game-day exercises with the on-call rotation.\n- Introduced service-level objectives for the six highest-traffic endpoints and wired the error budget into the release checklist, cutting rollbacks from 9 to 2 a quarter.\n- Rewrote the customer data export as a resumable job, taking the failure rate on exports over 2GB from roughly 30% to under 1%.\n- Ran the vendor evaluation for the observability stack and led the migration of 40 services onto OpenTelemetry tracing over two quarters.\n- Mentored 4 engineers, two of whom were promoted to senior within 18 months.\n\n### Software Engineer — Cobalt Systems (2019 - 2022)\n\n- Built an internal feature-flag service in Go, adopted by 14 teams and serving approximately 40,000 evaluations per second at peak.\n- Replaced a hand-rolled job scheduler with Temporal, retiring about 4,000 lines of bespoke retry logic and eliminating the duplicate-charge class of bugs.\n- Owned the PostgreSQL upgrade from 11 to 14 across six production clusters with zero customer-visible downtime.\n- Built the data retention tooling that brought the analytics warehouse into GDPR compliance ahead of the audit deadline.\n- Reduced CI pipeline duration from 26 minutes to 9 minutes by parallelizing the integration test suite and caching container layers.\n- On-call rotation lead for the platform team; wrote 11 of the team's runbooks.\n\n### Junior Software Engineer — Trellis Apps (2017 - 2019)\n\n- Implemented the REST API for the scheduling product used by roughly 300 clinics.\n- Built the appointment reminder service on Twilio, cutting no-show rates by 22% across the pilot group of 40 clinics.\n- Wrote the company's first integration test harness, raising coverage from 12% to 68%.\n- Migrated the frontend from jQuery to React over three months alongside feature work.\n\n### Software Engineering Intern — Puget Analytics (Summer 2016)\n\n- Built an internal dashboard in Flask and D3 that replaced a weekly manual spreadsheet report for the sales team.\n- Wrote the ETL job that backfilled three years of historical sales records.\n\n## Projects\n\n### pgslice-lite — open source (2021 - Present)\n\n- Postgres partition-management CLI in Go; 900+ GitHub stars, 23 outside contributors.\n- Wrote the migration planner that computes a zero-downtime cutover path for tables above 500GB.\n\n### Ledgerbook — side project (2020)\n\n- Double-entry bookkeeping (ledger) API in Python and FastAPI, used by two small businesses.\n\n## Technical Skills\n\n**Languages:** Go, Python, TypeScript, Rust, SQL\n**Data and messaging:** PostgreSQL, Kafka, Redis, Temporal, Airflow, dbt, Snowflake\n**Infrastructure and reliability:** AWS, GCP, Kubernetes, Docker, Terraform, OpenTelemetry, Grafana, GitHub Actions\n**APIs and frameworks:** REST, gRPC, FastAPI, Flask, React\n**Testing:** pytest, integration test harnesses\n\n## Education\n\n**B.S. Computer Science**, University of Washington, 2017\n- Minor in Statistics. Senior capstone: distributed rate limiter in Rust.\n- Teaching assistant for the undergraduate databases course, two semesters.\n\n## Certifications\n\n- AWS Certified Solutions Architect — Associate (2021, renewed 2024)\n- Certified Kubernetes Administrator (CKA), 2022\n\n## Talks and Writing\n\n- \"Paying Down Pipeline Debt\", PyCascades 2023 (conference talk).\n- Author of the Northwind engineering blog post on multi-region failover testing, approximately 30,000 reads.\n\n## Languages\n\nEnglish (native), Spanish (professional working proficiency)"
    },
    {
      "model": "claude-sonnet-5-5",
      "prompt": "rules",
      "resumeLines": 18,
      "keptWordForWord": 14,
      "inventedNumbers": 0,
      "postingEchoLines": 0,
      "driftFlags": 0,
      "driftFlagsThatHold": 0,
      "notWrittenItems": 9,
      "outputTokens": 5518,
      "batchCostUsd": 0.0309,
      "flaggedLines": [],
      "tailoredResume": "ALEX MORGAN\nSan Francisco, CA · alex.morgan.test@example.com · (415) 555-0142\nlinkedin.com/in/alexmorgantest · github.com/alexmorgantest · Authorized to work in the US\n\nSenior Software Engineer with backend services and reliability experience\n\nSUMMARY\n9 years building and operating backend services and developer platforms in Python,\nGo and TypeScript, with a focus on reliability, data pipelines and paying down\noperational toil.\n\nEXPERIENCE\n\nSenior Software Engineer — Northwind Data (2022 - Present)\n- Led the migration of the billing pipeline from a nightly batch job to a streaming\n  Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.\n- Cut p95 API latency from 820ms to 210ms by adding a read-through Redis cache and\n  removing three N+1 query patterns in the orders service.\n- Designed and shipped the multi-region failover runbook; led two live game-day\n  exercises with the on-call rotation.\n- Mentored 4 engineers, two of whom were promoted to senior within 18 months.\n- Introduced service-level objectives for the six highest-traffic endpoints and wired\n  the error budget into the release checklist, cutting rollbacks from 9 to 2 a quarter.\n- Ran the vendor evaluation for the observability stack and led the migration of 40\n  services onto OpenTelemetry tracing over two quarters.\n\nSoftware Engineer — Cobalt Systems (2019 - 2022)\n- Built an internal feature-flag service in Go, adopted by 14 teams and serving\n  approximately 40,000 evaluations per second at peak.\n- Owned the PostgreSQL upgrade from 11 to 14 across six production clusters with\n  zero customer-visible downtime.\n- Replaced a hand-rolled job scheduler with Temporal, retiring about 4,000 lines of\n  bespoke retry logic and eliminating the duplicate-charge class of bugs.\n- Reduced CI pipeline duration from 26 minutes to 9 minutes by parallelizing the\n  integration test suite and caching container layers.\n\nJunior Software Engineer — Trellis Apps (2017 - 2019)\n- Implemented the REST API for the scheduling product used by roughly 300 clinics.\n- Wrote the company's first integration test harness, raising coverage from 12% to 68%.\n\nPROJECTS\npgslice-lite — open source (2021 - Present)\n- Postgres partition-management CLI in Go; 900+ GitHub stars, 23 outside contributors.\n- Wrote the migration planner that computes a zero-downtime cutover path for tables\n  above 500GB.\n\nEDUCATION\nB.S. Computer Science, University of Washington, 2017\n\nCERTIFICATIONS\nAWS Certified Solutions Architect — Associate (2021, renewed 2024) · Certified\nKubernetes Administrator (CKA), 2022\n\nTECHNICAL SKILLS\nPython, Go, TypeScript, SQL, PostgreSQL, Redis, Kafka, Temporal, gRPC, Docker,\nKubernetes, Terraform, AWS, OpenTelemetry, pytest, GitHub Actions"
    },
    {
      "model": "claude-haiku-5-5",
      "prompt": "bare",
      "resumeLines": 34,
      "keptWordForWord": 23,
      "inventedNumbers": 0,
      "postingEchoLines": 0,
      "driftFlags": 2,
      "driftFlagsThatHold": 1,
      "notWrittenItems": null,
      "outputTokens": 3710,
      "batchCostUsd": 0.0011,
      "flaggedLines": [
        {
          "text": "Senior backend engineer with 9 years building large-scale distributed systems and REST APIs where correctness and reliability matter, from billing and payments pipelines to a feature-flag service handling about 40,000 evaluations per second at peak. Has led multi-quarter platform projects, cut production latency and operational risk, and mentored engineers into senior roles. Works across Go, Python, TypeScript and Rust.",
          "judge": "'large-scale distributed systems' is broader than sources ('backend services and developer platforms'); 'REST APIs' not in summary sources; 'billing and payments pipelines' adds payments domain beyond sources; 'feature-flag service handling about 40,000 evaluations per second at peak' — no feature-flag service or throughput figure in any source; 'cut production latency and operational risk' is a new summary claim; 'mentored engineers into senior roles' adds mentorship…"
        },
        {
          "text": "Distributed systems and data: Kafka, Temporal, Redis, PostgreSQL, Snowflake, Airflow, dbt, gRPC, REST API design",
          "judge": "'REST API design' as a named skill is not in vault or sources explicitly; sources show implementing a REST API but 'REST API design' as a skill label is vault-adjacent and the vault lists grpc but not REST explicitly. However this is borderline; the candidate implemented REST APIs so claiming REST API design is a reasonable restatement. Flagging because 'REST API design' as a discrete skill is not listed in the vault and goes slightly beyond what sources state."
        }
      ],
      "tailoredResume": "# Alex Morgan\n\nSan Francisco, CA · alex.morgan.test@example.com · (415) 555-0142\nlinkedin.com/in/alexmorgantest · github.com/alexmorgantest · Authorized to work in the US\n\n## Summary\n\nSenior backend engineer with 9 years building large-scale distributed systems and REST APIs where correctness and reliability matter, from billing and payments pipelines to a feature-flag service handling about 40,000 evaluations per second at peak. Has led multi-quarter platform projects, cut production latency and operational risk, and mentored engineers into senior roles. Works across Go, Python, TypeScript and Rust.\n\n## Experience\n\n### Senior Software Engineer, Northwind Data (2022 - Present)\n\n- Led the migration of the billing pipeline from a nightly batch job to a streaming Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.\n- Cut p95 API latency from 820ms to 210ms by adding a read-through Redis cache and removing three N+1 query patterns in the orders service.\n- Rewrote the customer data export as a resumable job, reducing the failure rate on exports over 2GB from roughly 30% to under 1%.\n- Introduced service-level objectives for the six highest-traffic endpoints and wired the error budget into the release checklist, cutting rollbacks from 9 to 2 a quarter.\n- Designed and shipped the multi-region failover runbook; led two live game-day exercises with the on-call rotation. Authored the Northwind engineering blog post on multi-region failover testing (about 30,000 reads).\n- Led the migration of 40 services onto OpenTelemetry tracing over two quarters, after running the vendor evaluation for the observability stack.\n- Mentored 4 engineers, two of whom were promoted to senior within 18 months.\n\n### Software Engineer, Cobalt Systems (2019 - 2022)\n\n- Built an internal feature-flag service in Go, adopted by 14 teams and serving approximately 40,000 evaluations per second at peak.\n- Replaced a hand-rolled job scheduler with Temporal, retiring about 4,000 lines of bespoke retry logic and eliminating the duplicate-charge class of bugs.\n- Owned the PostgreSQL upgrade from 11 to 14 across six production clusters with zero customer-visible downtime.\n- Reduced CI pipeline duration from 26 minutes to 9 minutes by parallelizing the integration test suite and caching container layers.\n- Built the data retention tooling that brought the analytics warehouse into GDPR compliance ahead of the audit deadline.\n- Served as on-call rotation lead for the platform team and wrote 11 of the team's runbooks.\n\n### Junior Software Engineer, Trellis Apps (2017 - 2019)\n\n- Implemented the REST API for the scheduling product used by roughly 300 clinics.\n- Built the appointment reminder service on Twilio, cutting no-show rates by 22% across the pilot group of 40 clinics.\n- Wrote the company's first integration test harness, raising coverage from 12% to 68%.\n- Migrated the frontend from jQuery to React over three months alongside feature work.\n\n### Software Engineering Intern, Puget Analytics (Summer 2016)\n\n- Built an internal dashboard in Flask and D3 that replaced a weekly manual spreadsheet report for the sales team.\n- Wrote the ETL job that backfilled three years of historical sales records.\n\n## Projects\n\n### pgslice-lite, open source (2021 - Present)\n\n- Postgres partition-management CLI in Go; 900+ GitHub stars, 23 outside contributors.\n- Wrote the migration planner that computes a zero-downtime cutover path for tables above 500GB.\n\n### Ledgerbook, side project (2020)\n\n- Double-entry ledger API in Python and FastAPI, used by two small businesses.\n\n## Education\n\n**B.S. Computer Science**, University of Washington, 2017\n- Minor in Statistics. Senior capstone: distributed rate limiter in Rust.\n- Teaching assistant for the undergraduate databases course, two semesters.\n\n## Certifications\n\n- AWS Certified Solutions Architect, Associate (2021, renewed 2024)\n- Certified Kubernetes Administrator (CKA), 2022\n\n## Talks and Writing\n\n- \"Paying Down Pipeline Debt,\" PyCascades 2023 (conference talk)\n- Northwind engineering blog post on multi-region failover testing (about 30,000 reads)\n\n## Technical Skills\n\n**Languages:** Go, Python, TypeScript, Rust, SQL\n**Distributed systems and data:** Kafka, Temporal, Redis, PostgreSQL, Snowflake, Airflow, dbt, gRPC, REST API design\n**Frameworks:** FastAPI, Flask, React\n**Infrastructure and reliability:** AWS, GCP, Kubernetes, Docker, Terraform, OpenTelemetry, Grafana, GitHub Actions, pytest\n\n## Languages\n\nEnglish (native), Spanish (professional working proficiency)"
    },
    {
      "model": "claude-haiku-5-5",
      "prompt": "rules",
      "resumeLines": 21,
      "keptWordForWord": 17,
      "inventedNumbers": 0,
      "postingEchoLines": 0,
      "driftFlags": 1,
      "driftFlagsThatHold": 1,
      "notWrittenItems": 12,
      "outputTokens": 11545,
      "batchCostUsd": 0.0031,
      "flaggedLines": [
        {
          "text": "Senior software engineer with 9 years building and operating backend services and developer platforms. Works across Python, Go and TypeScript, with a focus on reliability, low-latency distributed systems, data pipelines and paying down operational toil.",
          "judge": "Adds 'low-latency distributed systems' as a focus area; the sources list 'reliability, data pipelines and paying down operational toil' but never mention low-latency distributed systems as a focus."
        }
      ],
      "tailoredResume": "Alex Morgan\nSan Francisco, CA · alex.morgan.test@example.com · (415) 555-0142\nlinkedin.com/in/alexmorgantest · github.com/alexmorgantest · Authorized to work in the US\n\nSUMMARY\nSenior software engineer with 9 years building and operating backend services and developer platforms. Works across Python, Go and TypeScript, with a focus on reliability, low-latency distributed systems, data pipelines and paying down operational toil.\n\nEXPERIENCE\n\nSenior Software Engineer — Northwind Data (2022 - Present)\n- Cut p95 API latency from 820ms to 210ms by adding a read-through Redis cache and removing three N+1 query patterns in the orders service.\n- Led the migration of the billing pipeline from a nightly batch job to a streaming Kafka consumer, cutting invoice latency from 18 hours to under 4 minutes.\n- Introduced service-level objectives for the six highest-traffic endpoints and wired the error budget into the release checklist, cutting rollbacks from 9 to 2 a quarter.\n- Mentored 4 engineers, two of whom were promoted to senior within 18 months.\n- Rewrote the customer data export as a resumable job, taking the failure rate on exports over 2GB from roughly 30% to under 1%.\n- Ran the vendor evaluation for the observability stack and led the migration of 40 services onto OpenTelemetry tracing over two quarters.\n- Designed and shipped the multi-region failover runbook; led two live game-day exercises with the on-call rotation.\n\nSoftware Engineer — Cobalt Systems (2019 - 2022)\n- Built an internal feature-flag service in Go, adopted by 14 teams and serving approximately 40,000 evaluations per second at peak.\n- Owned the PostgreSQL upgrade from 11 to 14 across six production clusters with zero customer-visible downtime.\n- Replaced a hand-rolled job scheduler with Temporal, retiring about 4,000 lines of bespoke retry logic and eliminating the duplicate-charge class of bugs.\n- On-call rotation lead for the platform team; wrote 11 of the team's runbooks.\n\nJunior Software Engineer — Trellis Apps (2017 - 2019)\n- Implemented the REST API for the scheduling product used by roughly 300 clinics.\n\nPROJECTS\n\npgslice-lite — open source (2021 - Present)\n- Postgres partition-management CLI in Go; 900+ GitHub stars, 23 outside contributors.\n- Wrote the migration planner that computes a zero-downtime cutover path for tables above 500GB.\n\nLedgerbook — side project (2020)\n- Double-entry bookkeeping API in Python and FastAPI, used by two small businesses.\n\nEDUCATION\nB.S. Computer Science, University of Washington, 2017. Minor in Statistics. Senior capstone: distributed rate limiter in Rust.\n\nCERTIFICATIONS\nAWS Certified Solutions Architect — Associate (2021, renewed 2024); Certified Kubernetes Administrator (CKA), 2022\n\nWRITING\nAuthor of the Northwind engineering blog post on multi-region failover testing, approximately 30,000 reads.\n\nSKILLS\nPython, Go, TypeScript, Rust, SQL, PostgreSQL, Redis, Kafka, Temporal, Kubernetes, Docker, AWS, Terraform, gRPC, FastAPI, OpenTelemetry, GitHub Actions\n\nLANGUAGES\nEnglish (native), Spanish (professional working proficiency)"
    }
  ]
}