MTS - Research
Role summary by JobGrid
MTS - Research at interaction: Palo Alto, United States; On-site; Senior; IT; Data Science & ML. JobGrid adds normalized role facts, source context, and a path to the employer application page so candidates can compare the listing before applying.
- Location and workplace: Palo Alto, United States, On-site
- Role classification: IT, Data Science & ML, Senior
- Source freshness: checked by JobGrid on 2026-06-09.
- Application path: candidates continue to the employer application page with non-personal referral tags.
We’re the makers of Poke.com, a proactive AI agent for everyday life. Interaction is a $300M consumer company backed by $27M from General Catalyst and angels such as Guillermo Rauch (Vercel), Scott Wu (Cognition), Patrick and John Collison (Stripe), Fred Ehrsam (Coinbase), Ken Howery (Co-Founder of PayPal and Founders Fund), and many others.
The Context
For the last year, our core engineering team has built this entire agent platform from the ground up. Now, traffic and concurrent agent executions are scaling by orders of magnitude. We need engineers to harden infrastructure without slowing down feature velocity.
We only care about technical depth, obsession with your craft, and how well you can build alongside us. Our current team is strongly connected and talent-dense, with backgrounds from Jane Street, MIT/Stanford research, and International Olympiads, but we care more about experience building than your pedigree.
The Role
We hire across many strengths under one title. You are a T-shaped engineer who can navigate the whole stack, but you are uniquely experienced in at least one of these areas:
Scale & Infra: High-concurrency, low-latency, distributed state under heavy load.
Product: Prototyping, shipping, and iterating on agentic workflows zero-to-one.
Extensibility & DevTools: APIs, SDKs, and primitives that make Poke programmable by external developers.
AI Runtime & Evals: Tool use, model selection, and output quality in production.
Post-Training & Alignment. SFT, preference optimization, reward modeling, synthetic data generation, eval design, and improving agent performance through training rather than prompting.
We work mostly in Typescript and some Python, but lack of experience in these specific languages is not a deal breaker.
What we're looking for
Rigor under pressure. You move between infra, product, and debugging. You can stabilize melting systems while still shipping features.
Platform thinking. You design primitives, not one-off features. You build solutions that solve today's product needs while unlocking ten future use cases.
Low ego. The talent bar is high, but no important problem is beneath us.
Something Else
If you don't fit perfectly into a bucket, that’s fine. We care more about ownership and technical depth over specific labels.
Tell us what you are unusually good at, what systems you’ve owned, and why it transfers here. Include a link to your GitHub, a project you're proud of, or a breakdown of the hardest technical problem you've solved.