Senior Machine Learning Scientist
Remote ✓ENABOUT SWEATPALS Sweatpals is the community-first fitness platform turning workouts into social experiences. Backed by a16z speedrun, Patron, Kevin Hart, Pear VC, and founders of Instacart and Dreamworks Animations, we connect hundreds of thousands of "pals," hosts, and gyms through events, memberships, and social features. We're still scrappy at heart, but scaling fast. We believe working out should be joyful, social, and inclusive, not just a solo grind. From run clubs and beach pilates to pickleball leagues and cold plunge socials, Sweatpals turns everyday workouts into meaningful social experiences. Sweatpals also gives local leaders the tools to grow their fitness communities from side hustles to full-time, even million-dollar businesses. Hosts use our platform to run their business, from ticketing and memberships to marketing tools. We're an AI-forward company. If you're excited about working at the intersection of ML, product, and a real marketplace, you'll fit right in. THE ROLE We're looking for a Senior Machine Learning Scientist to join the AI Squad and own the most ambitious ML work on our roadmap. You'll report to the Head of AI and partner closely with engineering and product to ship models that move our marketplace. This is a high-ownership role. You'll define problems, run the science, ship to production, and measure real user impact. You won't inherit a graveyard of half-finished notebooks. You'll build the next layer of ML at Sweatpals on top of what we've already shipped: semantic search, event tagging, collection ranking, retention models, and our LLM-powered HostCopilot and Front Desk Agent. You'll spend your time on problems like: • How do we rank events for each Pal so they discover more hosts they'd love? • Can we predict churn early enough for HostCopilot to nudge before it happens? • What's the right way to price a class or membership to maximize host GMV without hurting bookings? • How do we make LLMs reliable enough to draft host campaigns, recommend events, and answer questions at a real front desk? • How do we measure if our models actually move the marketplace, not just CTR? WHAT YOU'LL DO MODELING & RESEARCH • Frame fuzzy product problems as ML problems and pick the right approach: ranking, retrieval, classification, sequence models, LLM agents, or classic stats • Run end-to-end: data exploration, offline evaluation, prototype, online experiment, iteration • Push to the cutting edge when it matters, stay pragmatic when it doesn't • Own offline metrics (NDCG, recall@k, AUC, calibration) and tie them to online metrics (booking lift, retention, GMV) PRODUCTION & SHIPPING • Ship models to production with our engineering team. Our ML stack is FastAPI, PostgreSQL, BigQuery, AWS App Runner, with retrieval via FAISS and sentence-transformers, and managed LLM APIs (Claude, Gemini) • Build evaluation harnesses and monitoring so we know when models drift • Keep latency budgets honest LLM & AGENT SYSTEMS • Develop LLM-powered features across HostCopilot (drip campaigns, retention nudges, pricing and content suggestions) and Pal-facing surfaces (AI Concierge, semantic search, recommendations) • Build agentic systems with tool use, RAG, structured outputs, evaluation loops, and human-in-the-loop where needed • Decide when to prompt-engineer, when to fine-tune, and when a classical model is the better answer CROSS-FUNCTIONAL IMPACT • Partner with product to size opportunities and translate findings into roadmap decisions • Partner with growth and our data analyst to measure marketplace impact rigorously • Set the bar for the squad on ML rigor: offline evaluation, experiment design, and writeups WHAT WE'RE LOOKING FOR EXPERIENCE • 5+ years of applied ML experience shipping models to production. Bonus if some of that was in marketplaces, search, or recommendations • Track record of taking a problem from "vague PM ask" to "shipped feature that moved a metric" • Comfort with the full lifecycle: framing, data, modeling, evaluation, deployment, monitoring TECHNICAL SKILLS • Strong Python and SQL. You write production code, not just notebooks • Solid foundations in at least one ML area: ranking and recommendation systems, NLP and embeddings, classical ML, LLMs and agents, or causal inference • Comfortable with modern LLM tooling: prompting, RAG, evaluation, tool use, structured outputs • Practical stats: experiment design, dealing with confounding, knowing when an A/B test is broken • Familiarity with our stack is a plus: FastAPI, PostgreSQL, BigQuery, FAISS, sentence-transformers, AWS, Amplitude • Advanced degree in ML, CS, stats, or a related field is typical. PhD or research background is a strong bonus MINDSET • Product first. You care about user impact more than novelty • You use AI tools daily. Claude Code, Cursor, whatever ships faster • You write things down. Memos, experiment results, design docs • …
Verantwortungsbereich:
MODELING & RESEARCH • Frame fuzzy product problems as ML problems and pick the right approach: ranking, retrieval, classification, sequence models, LLM agents, or classic stats
