Data Science Intern

Mercor•San Francisco, CA

26d•Onsite

About The Position

As a Data Science Intern at Mercor, you’ll join a fast-moving, metrics-driven engineering team that powers critical decisions across the company. You’ll analyze data that directly impacts ranking, hiring efficiency, candidate experience, and revenue. From day one, you’ll work with real datasets, ship insights used by product and engineering, and prototype models that improve how we match talent to AI companies. You’ll work closely with engineers, PMs, and leadership, designing experiments, evaluating LLM-powered systems, and building the foundations of data integrity and visibility across the platform. You’ll move quickly while maintaining a high bar for analytical rigor, clarity, and statistical correctness. At the end of the process, you’ll be team-matched to where you can have the most impact, on one of the following: Talent platform analytics – improving match quality, ranking, time-to-hire, and marketplace efficiency through experimentation and modeling. Applied AI/human data insights – partnering with leading AI labs (OpenAI, Anthropic, Google) to design evaluation rubrics, run human-in-the-loop studies, and understand how experts shape post-training data for frontier models.

Requirements

Pursuing a degree in a quantitative field (graduating 2025–2027).
Strong fundamentals in statistics, SQL, and Python.
Experience with experiment design, causal reasoning, and data analysis.
Ability to communicate clearly with engineers, product managers, and leadership.
Excited to work in person and thrive in a fast-paced environment.

Nice To Haves

Curiosity about LLM evaluation, retrieval, ranking, or marketplace dynamics (a plus).
experience with dbt, dashboarding tools, recommendation/search metrics, or LLM/agent evaluation.

Responsibilities

Defining north-star metrics and feature-level KPIs for ranking, interview analytics, and payouts systems.
Designing and running A/B tests and quasi-experiments; translating results into product decisions within days.
Building dashboards and lightweight data models that empower teams to self-serve insights.
Instrumenting events with engineers and improving data quality, observability, and latency.
Prototyping models (from baselines to gradient boosting) to improve matching and scoring systems.
Evaluating LLM-powered agents through rubric design, human-in-the-loop experiments, and guardrail canary testing.