Engineering leader for public systems. I build teams, portfolios, and the technical foundations those teams need to deliver.

Most recently, I was a Senior Director of Engineering at Coforma and one of three Senior Directors in its 50-person engineering organization. I led a 22-person reporting structure with five direct reports, including three engineering directors. I owned the company-wide healthcare engineering portfolio and served as engineering lead for business development.

That work stayed close to outcomes:

• Nine engineers moved into senior or leadership roles, including four promotions to Director or Principal. My teams had three voluntary departures across three years.
• I owned engineering for the seven-application CMS Medicaid & CHIP Data Collection Tools suite, oversaw Coforma's work on the federal Medicaid Drug Programs system, and led the firm's FHIR and health data interoperability specialization.
• I was engineering lead and principal engineer for MyCareer.NJ.gov, New Jersey's statewide workforce platform. It has served 1.8 million users since December 2023 and provides bilingual career, training, labor-market, tuition-assistance, and support-service tools.

On MyCareer.NJ.gov, I led the platform and set architecture across its three production codebases. The work included a zero-downtime GCP-to-AWS migration, test coverage raised from zero to 93–96%, a 94% reduction in known vulnerabilities, complete English and Spanish access, and the 2023 CTDL data modeling that began New Jersey's migration of its training program registry into the national Credential Engine ecosystem, where those programs are live today.

I also designed its shared applied-AI foundation and six primary proof-of-concept features. The foundation is now live in production; the features remained gated from end users pending pilot approval. The most mature was a bilingual, grounded career coach with cost-aware model routing, a merge-blocking adversarial test gate, and evaluation designed to expose weak categories rather than hide them. The 1.8 million-user count above applies to the production platform; the AI features were proofs of concept, not deployed services.

Before Coforma, I built public systems in California state government and public higher education:

• At the California Energy Commission, I led development of regulatory energy-data systems, helped move the commission's data infrastructure to AWS, and operated the Title 20 appliance certification system.
• At the California Public Utilities Commission, I directed the Transportation Carrier Portal, which launched in 2021, and architected a claims portal for telecommunications-access and broadband programs.
• At the California Department of Social Services, I built and operated caregiver-safety, licensing-payment, legislative-analysis, and benefits-program systems.
• At UC Berkeley's Graduate School of Education, I led application development for the equity-focused 21st Century California School Leadership Academy.

I hold an M.S. in Software Engineering from CSU Fullerton, which I completed while working full-time. I also hold a B.S. in Computer Science from the University of Oregon and an ITIL Foundation certification.

Based in Davis, California · Portfolio · LinkedIn

These independent projects are public and inspectable. Most are pre-1.0 tools or reference implementations, not evidence of organizational adoption. Each repository documents what has shipped, what remains experimental, and what still requires human review.

• tods-validate is a deterministic validator for the Transit Operational Data Standard. Version 0.8.0 is on PyPI, and the project ships a GitHub Action, container image, pre-commit hook, and browser playground.
• GTFS Scorecard (live at gtfsscorecard.org) is a transit-data-quality service covering a curated registry of 2,185 feed records, 2,182 of them with published scorecards as of August 2026; its status page reports the exact current counts. It also ships a GitHub Marketplace Action and a read-only MCP server.
• ctdl-validate is a deterministic structural validator for CTDL JSON-LD, the national credential-data standard: CTID grammar, identifier kinds, reference resolution, domain and range, and inverse consistency, with every finding cited to the published schema.
• disclosed grades US higher-education institutions on what they disclose rather than how they perform, across the complete 6,163-institution IPEDS directory and a 600-institution College Scorecard sample, with applicability rules that keep denominators honest.
• fhir-scorecard grades payer FHIR discovery endpoints daily under the CMS interoperability rules, every finding citing the specification text or the stated convention it applies, drift tracked between runs. Unauthenticated surfaces only, never patient data.
• Afterward (formerly Camino, live at afterward.chelseakr.com) joins 3,266 California training programs to their federally reported outcomes and the state's occupation and wage projections, built entirely from public data. More than a third publish no outcome data, and the site says so rather than rendering an absence as a zero. No account, no tracking, English and Spanish from the first release.
• NearMiss is a beta road-safety analysis toolkit that uses exposure-normalized rates, confidence intervals, and statistically controlled hotspot detection instead of treating raw report density as risk.
• Swelter (live demo, refreshed daily on real data) is a beta reference system for community heat and air quality, with correction provenance, a bilingual dashboard, and OGC SensorThings exports.
• ledger is a beta, privacy-first digital-preservation tool using BagIt, PREMIS, Dublin Core, encrypted contributor identities, and consent-based disclosure. It contains synthetic and consented fixtures only.
• outcome-receipts is a beta tool for nonprofit funder reports where every figure carries a receipt: a deterministic query, a data-slice hash, and a fail-closed grounding gate that blocks export if a number cannot be traced to evidence.
• habitable is an alpha, offline-first tool for tenant unions that makes habitability evidence tamper-evident with content hashes and RFC 3161 timestamps, then syncs peer-to-peer over an end-to-end-encrypted CRDT; an optional relay only ever carries ciphertext.

I lead at portfolio altitude and stay technically deep: I set architecture, challenge assumptions, and understand delivery risk. Grit is not a staffing model: I build clear ownership, strong managers, useful standards, and systems that do not depend on one heroic person.

My current technical center of gravity is TypeScript, Python, React and Next.js, AWS, PostgreSQL, data interoperability, applied AI evaluation, and delivery systems. Earlier work also spans C# and .NET, Salesforce, Snowflake, Azure, GCP, and Natural/ADABAS modernization.

I treat accessibility, privacy, security, operability, and multilingual delivery as engineering requirements. For AI systems, that means evidence-constrained generation, explicit refusal, adversarial testing, disaggregated evaluation, cost controls, feature flags, and human review for legal, policy, subject-matter, and accessibility judgments.

AI agents are part of my development workflow. I choose the architecture, write the acceptance criteria, review the output, and decide whether it is ready to release.

• I won't work on weapons, warfare, policing, mass surveillance, or technology that profits from incarceration.
• I look for organizations whose work helps people routinely failed by public systems, and whose leadership reflects the communities they serve.
• I won't use AI to decide whether someone gets a job, benefit, service, or opportunity. It can support a person's decision, but that person should be able to inspect the evidence, correct bad information, and make the final call.
• I collect as little personal data as practical and prefer local or offline tools when they fit.
• I separate sources, calculations, and interpretation. When a system does not know, it should say so.

In 2026, the Labor Market Information Institute gave NJDOL its Best State LMI Focus on Impact and Sustainability award for the department's evidence-based approach and MyCareer.NJ.gov. The platform has been featured by Fast Company and the Royal Statistical Society's Real World Data Science. The Harvard Kennedy School's Project on Workforce described its Training Explorer as a "promising potential model," and Credential Engine has featured its CTDL publishing work. The platform was also nominated for a NASWA Workforce Innovation Award. I was a registered member of Credential Engine's CTDL Advisory Group from January 2025 to July 2026, and in 2023 I joined a Digital Services Coalition roundtable on AI and machine learning in federal and civic technology.

I'm interested in engineering leadership roles (VP of Engineering, Head of Engineering, Senior Director, Director, or Principal Engineering Manager) and independent consulting engagements in the same domains while I search. I want to lead teams and managers, shape architecture and delivery systems, and build reliable, accessible technology in public health, workforce and social services, energy and utilities, government digital services, or responsible AI.

Reach me at chelseakr.com or on LinkedIn.