Data scientist

Companion Labs is an early-stage, AI-native product company building from Bengaluru. The team is at the Seed stage — small, deliberate, and moving fast toward a product that learns from its users in ways most software still doesn't.

The founding data hire at Companion Labs. You will own the data layer before anyone else does — the instrumentation, the models, the analytical frameworks — and what you build in the first six months will define how this team makes decisions for years.

This is an IC seat at a Seed-stage company. There is no existing data infrastructure to inherit, no backlog of dashboards to maintain, and no team to hand work off to. You build it, you own it, and you earn the right to shape what comes next.

• The instrumentation layer. Define what the product tracks, how events are structured, and what questions the data will be able to answer — before the first user cohort scales.
• The analytical framework. Build the metrics that matter: retention, engagement, behavioral signals. Not a dashboard library — a decision-making system.
• The first production models. Own the full arc from problem framing to deployment. You ship models that change how the product behaves, not notebooks that inform a meeting.
• The data infrastructure. Pipelines, warehousing, tooling choices — you make the calls that the next data hire will inherit. Build it like it has to last.
• Product insight delivery. Translate findings into recommendations the founding team acts on. A result that sits in a doc is a result that didn't ship.
• The experimentation framework. Design and run A/B tests and behavioral experiments. Own the methodology so the team can run experiments without you as a bottleneck.
• The data roadmap. As the company scales, you define what the data function becomes — what to hire for, what to build next, what to stop doing.

• You've built a data function from scratch — not joined one mid-flight. You know what it feels like to instrument a product before anyone else has, and you've made the calls on what to track and why.
• You're comfortable being the only data person in the room for an extended stretch. No senior data lead to escalate to, no peer to sanity-check with. You've operated that way before and you preferred it.
• You've worked on a consumer or B2C product where user behavior was the primary signal. You know how to read engagement curves, retention cohorts, and drop-off patterns and turn them into a product decision.
• You write SQL the way a good engineer writes code — clean, documented, and built to be reused. You don't need a data engineer to translate your questions into queries.
• You've shipped at least one model into production. Not a notebook, not a proof of concept — something that ran in a live product and changed how it behaved.
• You have a point of view on what AI-native products should measure that legacy analytics frameworks miss. You've thought about this, and you can defend it.
• You're drawn to early-stage environments because the ambiguity is the interesting part, not despite it. You've been in a 0-to-1 setting before and you know what it costs.
• You move between analysis and action without needing a handoff. A finding doesn't sit in a doc — it becomes a recommendation, and you follow it through.
• You can explain a model's output to a founder who doesn't care about the math. Clarity over precision when the audience demands it.

• Own the outcome, not the task. Finishing the analysis is the start. What changes because of it is the job.
• Write it down. Decisions, assumptions, dead ends — documented so the team doesn't re-litigate them six months later.
• Move at product speed. A good-enough answer today beats a perfect one next week. Know when to ship the insight and when to go deeper.
• Build for the next person. Every pipeline, every model, every dashboard is infrastructure. Write it like someone else will maintain it.
• Disagree in the room. If the data says something the team doesn't want to hear, say it clearly and early. Silence is not diplomacy here.
• Stay close to the product. Data work that drifts from the user problem becomes overhead. Keep the question in view.
• Default to action. Ambiguity is not a reason to wait. Make a call, document the assumption, and move.

Seed-funded and building in Bengaluru. Compensation ranges from 15L to 30L, with equity on a founding-hire basis. This is the seat where the data function gets defined — the frameworks you build here will run the company's decision-making for years.