Snaphomz is an AI-first homebuying platform operating in California and Texas two of the highest-volume residential real estate markets in the country. We’re pre-Series A, incorporated in Delaware, with a working product and real users. We’re building fast.
The Role:
We’re looking for a PhD in machine learning, NLP, or information retrieval who is done with — or nearly done with — their program, and ready to work on problems that are actually hard.
This is a full-time research role with production responsibility. You will design and run experiments, build training pipelines, and ship models that real users interact with. There is no handoff between research and engineering here. You own it end to end.
We don’t hire researchers to write internal reports. We hire them to build things that work.
What You’ll Work On:
The problems in this domain are genuinely unsolved. User intent in property search is noisy, multi-signal, and shifts over time. Outcome labeling is ambiguous. Latency constraints are tight. Ground truth is delayed by months.
You’ll work across:
- Domain-specific model fine-tuning and continuous learning pipelines
- Semantic understanding for search — going well beyond what the user typed
- Behavioral modeling from interaction data
- Evaluation infrastructure that actually tells us when things get better
The specific roadmap is something we share with candidates we’re serious about. What we can say publicly: the research agenda here is not incremental. We’re not optimizing existing benchmarks. We’re working on things that haven’t been done in this space.
What We Need:
The fundamentals. Fine-tuning — LoRA, QLoRA, full — and when each is the right call. Dense and sparse retrieval, reranking, hybrid search. Evaluation design on tasks where ground truth is noisy or delayed. You need to know this work, not just know of it.
Speed. We move fast. The gap between “interesting direction” and “working prototype” should be days for you, not months. If you need a lot of process to do good research, this isn’t the right environment.
Production instinct. You don’t need to be a software engineer. You do need to be able to get a model serving in a real environment. Notebooks are a starting point, not a finish line.
Depth over breadth. We’d rather you know one area of this field exceptionally well than have surface familiarity with all of it. Strong fundamentals travel. Shallow exposure doesn’t.
Experience in conversational AI, dialogue systems, or behavioral modeling is a strong signal. Real estate domain knowledge is useful but not a requirement — the domain can be learned quickly with the right foundation.
Research and Publication
We will publish. When we’ve built something worth saying, you’ll have the full support of the team to write it and put it out. We take this seriously — not as a credential exercise, but because doing work rigorous enough to publish is a standard we want to hold ourselves to.
What You’ll Get
- A research mandate with real scope — this is not a research-flavored engineering position
- Direct access to the founding team from day one
- Equity commensurate with how foundational this role is
- Competitive compensation — we’ll be straightforward about numbers in the first conversation
- A publication roadmap that’s treated as a first-class company commitment
How to Apply
Send your CV to careers@snaphomz.com and two or three sentences on what area of your PhD work you’re most proud of and why. That’s it. We’ll take it from there.
We review applications on a rolling basis and respond to every one.
No recruiters. No agencies.
Snaphomz is incorporated in Delaware and operates across the United States. We are an equal opportunity employer.