DS/ML Intern

Job description

A builder mindset is the core of this role and where you'll spend most of your time. But we're a small team building a whole product, not a research lab. The best person here treats ML systems as their primary craft while staying willing to do whatever the product needs — thinking through the product itself, shipping backend or frontend code, untangling data pipelines. We're looking for someone energized by the breadth, not someone who wants to stay in their lane. **What you'll work on** Evaluation systems for AI features - Help build the eval backbone our AI features ship against — failure taxonomies, LLM-as-judge rubrics, golden datasets, calibration against human judgment. - Learn what it takes to keep automated scores honest as models and prompts change. A feature with no eval has no quality floor. Model routing & inference economics - Get hands-on with how we route work across models — balancing cost, quality, and latency per task. - Help run the experiments that justify those choices and catch regressions. Scoring, measurement & signal quality - Work on turning noisy, real-world signals into scores you can actually trust — grounded in real statistical rigor, not vibes. - Help move heuristic-driven approaches toward calibrated, monitored systems. MLOps & production - Get exposure to the full lifecycle — feature pipelines, model versioning, rollout, monitoring for drift and silent quality decay. - Work alongside engineering to see how models get served reliably at low latency. **What we're looking for** Must have - Currently pursuing or recently completed a degree in CS, DS, ML, or a related field. - Some hands-on DS/ML experience — coursework, personal projects, research, or a prior internship — where you've built and run something end to end, not just notebooks. - Comfort with Python and working SQL knowledge. - Basic grounding in applied statistics — you can explain what a metric means and when it might be misleading. - A builder's instinct — genuinely curious about product decisions, backend, or frontend, not just the modeling layer. - Some exposure to LLMs — prompting, using APIs, or experimenting with model behavior. Nice to have - Any exposure to evaluation or observability tooling for LLM features. - Coursework or projects in information retrieval, entity-matching, or record-linkage. - Interest in developer-productivity, code analytics, or DevEx data.