Senior Research Scientist

About Firsthand Firsthand has built the first AI-powered Brand Agent platform, transforming the way marketers and publishers engage consumers through their own AI agents, anywhere online. While most AI applications in marketing and advertising focus on back-office automation, the Firsthand Brand Agent Platform™ powers front-line consumer engagement. Operating across both owned properties and paid media, Firsthand's Brand Agents make a company’s expertise accessible in real time, adapting to consumers’ interests and guiding them towards the information they need to take action. Central to the platform is Lakebed™, the company’s AI-first data and knowledge rights management system that ensures brands retain full ownership and control of their expertise. Firsthand is led by Jon Heller, Michael Rubenstein, and Wei Wei, whose previous ventures helped build the foundations of modern digital advertising. Backed by Radical Ventures, FirstMark Capital, Aperiam Ventures, and Crossbeam Venture Partners, Firsthand is shaping the future of AI-driven consumer engagement. Firsthand is headquartered in NYC, with team members working together in-office three days a week. About the Role We're looking for an applied research scientist to join our small, high-leverage research team. You'll prototype new AI-powered product capabilities, architect agentic workflows that connect LLM reasoning to production infrastructure, and build the evaluation frameworks needed to trust and ship those systems. You'll work closely with engineering and product — and your work will have a direct impact on what we build and how fast we build it. This role is hybrid (3 days/week in office), with a strong preference for candidates based in New York. Key Responsibilities New Capability Prototyping Research and build novel AI-powered features (e.g., image recommendation, personalization) using frontier models Move from ambiguous research questions to working demos fast — You value practical, timely results over slow, overly-refined outputs. Agentic System Design Design and implement agentic workflows across a variety of patterns, from single-agent tool use to multi-step orchestration to human-in-the-loop systems Build systems where agents can take meaningful actions with appropriate guardrails, not just generate text Evaluation Methodology Develop evals for agentic systems: correctness, reliability, latency, and behavior under edge cases Create frameworks the broader team can use, not one-off scripts Cross-functional Collaboration Translate research outputs into artifacts that product and engineering can understand and act on Partner with engineering and product to identify high-leverage capability gaps Who You Are Must-haves 4+ years of experience in applied ML/AI research or applied science roles 6+ months of hands-on agent-building experience (prompt chaining, tool use, multi-step reasoning, evaluation) Strong proficiency with frontier model APIs and coding agents (e.g., Claude Code, Codex, OpenCode, or equivalent), comfort with prompt engineering, tool use, and agentic patterns required Track record of taking ambiguous problems from first principles to working implementation Strong problem abstraction skills, can decompose an ambiguous business problem into a clear AI research formulation Proficiency in Python; comfortable writing production-quality code end-to-end Nice-to-haves Experience in AdTech or MarTech Experience with marketing measurement (attribution, incrementality, experiment design) Experience with LLM and agent eval frameworks Prior work at a startup or in a small team where you owned research end-to-end How to Apply If you are ready to embark on an exhilarating journey at the forefront of AI, seize this incredible opportunity and apply here. We eagerly anticipate hearing from you! Note: Compensation and equity will be market-competitive for well-capitalized, early stage startups and will be discussed during the interview process. #J-18808-Ljbffr