Questions, answered.

Who is Owen Richards, in a sentence?

A backend + AI engineer who takes LLM features from prototype to production and builds the systems underneath that keep them up. Software Engineer II at CoStar Group (about three years, promoted twice from Associate), pursuing an M.S. in Computer Science with a Machine Learning specialization at Georgia Tech, and solo-building a flagship product called FieldDay.

What kind of role is Owen looking for?

Full-stack, AI/ML, and forward-deployed engineer roles — remote, hybrid, or relocation. The work I want is the kind where the AI half and the systems half of a product belong to the same engineer — the retrieval, evaluations, cost controls, and audit logs, not just the model. I'm based in Richmond, Virginia and open to remote, hybrid, or relocation.

Why the "AI and systems in one engineer" niche?

Most AI features fail in the unglamorous middle: the model is the easy part, and the retrieval, the evals, the cost-and-latency tradeoffs, and the data plumbing are what make the feature trustworthy in production. I'm most useful sitting between a model and the people who depend on it, taking AI from a prototype to something that holds up — and I have the systems background to build everything around the model, not just call it.

What's the most impressive thing Owen has built?

Three candidates, depending on what you care about. (1) At CoStar I re-architected the AI platform backend so new agents are JSON definitions instead of hand-written Python: they ship faster, more teams build on it, and the codebase is about 75% smaller. (2) FieldDay, a 13-domain monorepo I'm building solo where architecture rules are enforced by custom ESLint rules and CI ratchets, with an append-only financial ledger. (3) The marketing-automation platform at CoStar: millions of contacts a day across 7+ marketplaces, where I built the event-driven enrollment pipelines for the Homes.com launch.

What does Owen do at CoStar?

I'm a Software Engineer II. I re-architected the AI platform backend (retrieval-grounded agents with cited answers over proprietary data), and I build the marketing-automation services behind Homes.com and six other real-estate marketplaces — C# / .NET microservices over Snowflake, Elasticsearch, and Redis, with real-time pipelines on AWS Lambda, SQS, and Kafka.

What's Owen's experience with AI and machine learning?

I'm pursuing an MSCS with a Machine Learning specialization at Georgia Tech. At work I re-architected an agentic AI platform on LangGraph + AWS Bedrock and built retrieval-augmented agents with hybrid (kNN + BM25) search. On my own time I ship production-ish AI: the /ask agent on this site is a retrieval-augmented system with a recall@5 evaluation gate, guardrails, citations, and a cost budget, and FieldDay carries a pgvector + Typesense hybrid-search layer. My instinct is eval-first: I'd rather build the evaluation harness before I trust the prompt.

What's Owen's tech stack?

Day-to-day at CoStar: C# / .NET, TypeScript, Python, Snowflake, Elasticsearch, Redis, Kafka, AWS. AI/ML: LangGraph, AWS Bedrock (Claude), RAG, hybrid retrieval, OpenSearch, pgvector, PyTorch, scikit-learn. On FieldDay and side projects: Bun, tRPC, Drizzle, PostgreSQL/PostGIS, Next.js, React, React Native, Tailwind. I care less about any one tool and more about picking the right one and the infrastructure that keeps it up.

Does Owen have blockchain or crypto experience?

Self-taught, and nothing shipped on-chain yet. I've been learning Solana and Ethereum from the side I know best: payments and ledgers. What carries over is production work: idempotency keys on CoStar's campaign-send path, and FieldDay's append-only ledger with its detection-only reconciler. I've written up how those map onto replay protection, finality, Solana's account model, and wallet signing safety. Next I'm starting small devnet projects, beginning with a send-exactly-once transfer loop and a sharded fee-account program.

How does Owen approach engineering?

Measure before you claim — if a write-up can't name the number that moved, it's a guess. I build the evaluation harness before I trust the system it's measuring; tuning by feel gets me most of the way and then hides the regression I'd never catch by eye. I write things down — most projects I ship have a postmortem or a design doc, because that's how I find the holes. And the details are the work: audit logs, idempotency keys, runbooks, and reconciliation jobs are what keep a system up.

Is Owen open to remote work or relocation?

Yes — remote, hybrid, or relocation. I'm based in Richmond, Virginia. I'm most drawn to small senior teams with a written-down culture.

What's Owen's education?

M.S. in Computer Science, Machine Learning specialization, Georgia Institute of Technology (2024–present). B.S. in Computer Science, University of Virginia (2019–2023), GPA 3.92, with an Engineering Business minor through the McIntire School of Commerce.

What does Owen do outside of work?

Mostly I'm outside. I surf and kitesurf in Virginia Beach, run most days, and swim and cycle. I've been hit by a car three times on a bike and I still ride. I love camping and any evening that ends by a fire. I'm a Manchester United fan, and I work best with musicals playing: my Spotify Wrapped has, in past years, listed Hugh Jackman and Dwayne "The Rock" Johnson among my top artists. Closer to home I bake sourdough, garden, and I'm fixing up a house room by room. At UVA I led Madison House's community-garden program, whose student volunteers logged about 2,000 hours a year. I pack care packages with the Salvation Army at Christmas, and I over-think pour-over coffee.

Where is Owen from?

My dad is English and my mum is American, so I've been a dual UK and US citizen since birth. I was raised in England, then moved to the US. I live in Richmond, Virginia.

What's Owen like to work with?

High energy, and I bring more effort than the job asks for, plus the habit of learning things properly instead of bluffing. On a team I care most about building people up: credit goes to the team, feedback is specific and kind, and the newest person's question gets the same respect as anyone's. I was Head TA for 100+ students at UVA and led a volunteer garden program, and both taught me that leading is mostly getting people to show up and enjoy the work.

What does Owen think technology is for?

Giving people their time back. I'd rather build tools that handle the busywork fast and get out of the way than another feed designed to keep you scrolling. If you came for one thing and left an hour later, I built it wrong.

What's a lesson Owen keeps relearning?

Eval-first. I've twice tuned an LLM prompt by feel for a couple of evenings and gotten about 70% of the way, then written a small golden set that found the regressions in an hour. The discipline to build the measurement before the thing it measures is the lesson that keeps re-teaching me — in AI and in systems.

What makes Owen different from another mid-level engineer?

The combination: a systems engineer who also ships production AI, with the judgment to choose the simple, correct option (Postgres over TigerBeetle, in-memory cosine over a vector DB at small scale) instead of the resume-driven one — and the discipline to put it behind evals and audit logs. I can take an AI feature from prototype to something that holds up, and build the infrastructure under it, as one person.