Data Analyst - StefanBrain

Turn StefanBrain's product, user, and LLM signal into decisions the whole team acts on.Role SummaryStefanBrain is an AI-powered marketing platform used by 80+ DTC businesses spending over $100M/month combined on digital advertising. We are building the data backbone that powers automated performance analysis, creative intelligence, and — increasingly — autonomous ad optimization for those brands.We’re hiring a Data Analyst who operates like a hybrid of a product owner and an engineer. You own the answer to “what are users actually doing, why, and what should we do about it” — and you ship the dashboards, instrumentation, LLM analyses, and chat-mining pipelines that make the answer visible to the whole team.We build AI-native. Claude Code, Cursor, Codex, and similar tools are part of the daily stack — not a side experiment. If you’d rather hand-code everything because that’s how you were trained, this isn’t the right fit.Role at a Glance$75K – $150KRemote (global)  ·  Overlap with the 7:00–11:00 PM CET (1:00–5:00 PM EST) windowJOB TYPEFull-timeENGAGEMENTIndependent contract to startEXPERIENCE2–5 yearsREPORTS TOFounder & Technical LeadSKILLSSQL  ·  Python  ·  PostHog / Amplitude  ·  dbt  ·  LLM observability (Langfuse / LangSmith)  ·  Chat & text mining  ·  Embeddings  ·  Funnels & cohorts  ·  A/B testingWhy You Should ApplyAt most companies, the Analyst role is "make a dashboard for the VP." Here, you own the read on every meaningful product question for a platform 80+ DTC brands use to spend $100M+ a month, and you propose what we build next. You'll work directly with the founder and Technical Lead, push product bets through sprint planning, and watch your recommendations ship the same week instead of dying in a quarterly roadmap review. The parent company is profitable. The product has PMF. The team is small enough that no one needs to remind you that your work matters. And the AI tooling - Claude Code, Cursor, Codex, LLM-assisted chat-mining loops - is your daily stack, not a side hobby.What You’ll DoAnalysis and product judgment.  Impact: this is how the company stops making product calls on gut and starts making them on signal.  Own the recurring read on product, user, and business behavior — usage across every tool, tied to activation, retention, expansion, churn, and acquisition efficiency. Define KPIs, present in sprint reviews, propose what we should build next. The team’s go-to read on every meaningful product question.Product analytics + LLM observability.  Impact: this is how we know whether the product is actually working — both for users and for the LLM under the hood.  Own PostHog or Amplitude end to end (taxonomy, funnels, cohorts, feature flags). Stand up the LLM observability surface (PostHog LLM, Langfuse, Helicone, LangSmith, or custom) — tokens, latency, cost, model and prompt-version performance, hallucination rates, prompt and model A/B tests. Partner with the Technical Lead on benchmarking.Chat data mining.  Impact: this is how a million chat turns turn into a product roadmap.  Cluster intents, surface failure modes, track week-over-week shifts. Turn unstructured conversation data into concrete product recommendations.Dashboards and the analyst-side modeling layer.  Impact: this is the surface the team makes calls off every day — it has to be right and it has to be alive.  Build the custom dashboards we run off. Own the dbt mart / semantic layer on top of the engineer’s raw tables, dashboard health, and PostHog event-taxonomy hygiene.SECONDARY RESPONSIBILITIESDocument every dashboard, metric, and analysis so teammates can use them without routing through you. Bring strong opinions on event schemas and analytics modeling — the engineer makes the architectural call, your job is to make sure analytics needs are heard. Onboard new data sources for analysis as connectors and product features come online. Treat improving your own AI-assisted workflow (Claude Code skills, custom prompts, agentic chat-mining loops, LLM-written briefings) as part of the job, not a side hobby.What We’re Looking ForMUST-HAVES2–5 years in analytics, product analytics, growth, or BI:  at a real product company — not a reports-on-request shop.Fluent in SQL and Python:  Daily, production-grade. You write transformation code and analysis scripts, not just queries.Have shipped custom dashboards that real teams used and acted on:  Show specific examples — what you built, who used it, what changed because of it.Hands-on ownership of PostHog or Amplitude:  Set one up from scratch or substantially rebuilt one — taxonomy, instrumentation, funnels, cohorts, retention, feature flags. You know what breaks when event schemas drift.LLM / AI product analytics experience:  Prompts, completions, agent traces, model comparisons. Either you’ve built analytics around an LLM product, or you’ve gone deep enough on LangSmith / Langfuse / Helicone / PostHog LLM / eval frameworks to be effective day one.Chat / text mining at scale:  Clustering, topic modeling, embeddings, LLM-assisted classification. Turn millions of chat turns into actionable themes.Product-owner mindset:  Bring opinions on what should change, propose product bets backed by data, push them through sprint planning. You influence the roadmap with sharp insight rather than wait for briefs.Analysis depth:  Take a vague business question (“are we losing power users? why?”), turn it into a sharp investigation, and return with an answer the team acts on.STRONG-TO-HAVESDTC / direct-response / paid-acquisition background:  funnels, CAC, ROAS, hooks, angles, creative testing.LLM observability in production:  LangSmith, Langfuse, Helicone, PostHog LLM, or rolled your own.Comfort with ad platform API data:  Meta, Google Ads, TikTok.Vector / embeddings experience:  for chat search, semantic clustering, LLM-assisted classification.LLM-as-judge evals:  or other automated quality metrics for AI products.CompensationIndependent contractor engagement. Monthly, paid in USD. Varies by region and experience. You’ll share your preferred comp when applying.