Junior Generative AI Engineer

Cleary Gottlieb Steen & Hamilton LLP is a global law firm with an internal team that builds bespoke AI solutions for legal work. The Junior Generative AI Engineer will help prototype, build, test, and deploy LLM-powered products for analyzing complex legal documents, while collaborating with engineers, data scientists, lawyers, and product specialists.

Responsibilities:

• Contribute to production AI systems: Under the guidance of senior engineers, help build, test, and improve LLM-powered document analysis features (extraction, classification, risk flagging) that serve lawyers daily
• Data Preparation & Engineering: Help clean, structure, and label legal data to power our AI systems — building the data pipelines that make models work well in practice
• Build features within agent workflows: Implement bounded components of multi-step AI workflows for legal tasks (e.g. due diligence, contract review) — writing tool integrations, prompt chains, and retrieval logic under the supervision of senior engineers
• Support evaluation, quality, and performance: Write test cases, contribute to automated eval pipelines, and help maintain regression suites. Gain exposure to model selection, token budgets, caching strategies, and the guardrails (output filters, PII redaction) that keep production AI systems reliable
• Collaborate with lawyers, engineers, and product: Participate in cross-functional sessions to understand legal workflows, contribute to acceptance criteria, and iterate based on user feedback and evaluation results
• Dedicated mentor: A senior engineer assigned as your day-to-day mentor for at least your first six months — available for architecture discussions, code review, and career guidance
• Structured onboarding: A 90-day onboarding plan covering our stack, legal domain fundamentals, and AI evaluation practices so you're never left guessing what to learn next
• Pair programming: Regular pair-programming sessions with senior engineers on production features — the fastest way to absorb patterns, tooling, and judgement
• Code review as learning: Every pull request receives a thorough, constructive review. You'll also review others' code early on, reading good code is one of the best ways to write it
• AI safety and bias training: Guided training on identifying non-deterministic failure modes, prompt injection risks, bias in LLM outputs, and the responsible-AI practices that govern our systems
• Growth path: A clear progression framework from Junior to Mid-Level to Senior, with regular check-ins, stretch goals, and the opportunity to take on increasing ownership as your skills develop