Staff Machine Learning Engineer

Primer exists to make the world a safer place. We build trusted, decision-ready AI for the organizations that can least afford to be wrong: the leaders, operators, and analysts making high-stakes calls about a world that changes faster than anyone can read it. We have offices in San Francisco, Pasadena, CA, and Arlington, VA. Learn more at primer.ai. The world produces more information every day than any team can read. We build the AI that turns it into a decision you can trust: fast and grounded enough to stake real consequences on.
As a Staff Machine Learning Engineer, you’ll own AI-driven products end to end, from the prototype that proves an idea works to the production system the mission depends on. You work fluently across the modern stack: large language models, agentic systems, retrieval, embeddings, fine-tuned models, and the evaluation harnesses that catch them when they drift. You can move fast on an open-ended problem and then harden the result into something reliable at scale, drawing on real distributed-systems experience. You’re energized by what frontier models and agents make newly possible. You reach for them reflexively, including to multiply your own output, and you raise the level of everyone around you as you go.
Your technical range matters as much as your judgment about how the pieces fit together. The best work here is designed as part of the whole system, not bolted on beside it, and the engineers with the most impact bring product managers and teammates along with them rather than around them. You’ll partner with product, engineering, and other technical leaders on the hardest version of the problem: putting real AI in the hands of people the moment a decision can’t wait.
Role and Responsibilities - How You Will Make an Impact

• Set both technical and product direction, making the strategic bets across our LLM, agentic, and NLP systems while shaping how we design AI-driven experiences and set expectations with users.
• Design and build the distributed, agentic systems behind our products at company-wide scale: tool-using conversational agents, multi-turn context, retrieval-grounded reasoning, and the orchestration that ties them together.
• Turn massive, messy, real-world data into the trustworthy signal agents' reason over: entity recognition and linking, relation extraction, summarization, semantic search, and knowledge-graph generation.
• Take models from checkpoint to production: package, deploy, and operate low-latency, high-concurrency inference (Triton, vLLM, GPU-backed serving) that stays fast and reliable under real load.
• Build the evals and labeled-data flywheels that steer the work rather than gate it, turning “it feels better” into proof you can act on.
• Partner with and influence cross-functional teams to shape the technical roadmap, and drive issues to root cause when quality or operations are on the line.
• Raise the engineering bar with better patterns, better practices, and a standard other engineers want to match.

Relevant Skills and Experience

• BS, MS, or PhD in computer science, a related field, or equivalent practical experience.
• 6+ years building production backend software, with a track record of shipping and operating ML-driven functionality.
• Mastery of data structures and algorithms, and the judgment to translate user needs into practical solutions.
• Hands-on depth with LLMs and agentic systems (prompt and context engineering, tool use, retrieval and RAG) and the broader ML toolkit such as PyTorch, plus experience defining evals to measure and improve quality.
• Fluency authoring production APIs in Python (Rust a plus).
• High agency and a bias to action in ambiguous, fast-moving problems, plus the curiosity, generosity, and love of teaching that lifts a whole team.

Bonus Points

• Experience with knowledge graphs, information retrieval at scale, or LLM fine-tuning and post-training.
• A pull toward high-stakes, real-world problems where the work actually ships and someone depends on the answer.

Role and Responsibilities - How You Will Make an Impact

• Set both technical and product direction, making the strategic bets across our LLM, agentic, and NLP systems while shaping how we design AI-driven experiences and set expectations with users.
• Design and build the distributed, agentic systems behind our products at company-wide scale: tool-using conversational agents, multi-turn context, retrieval-grounded reasoning, and the orchestration that ties them together.
• Turn massive, messy, real-world data into the trustworthy signal agents' reason over: entity recognition and linking, relation extraction, summarization, semantic search, and knowledge-graph generation.
• Take models from checkpoint to production: package, deploy, and operate low-latency, high-concurrency inference (Triton, vLLM, GPU-backed serving) that stays fast and reliable under real load.
• Build the evals and labeled-data flywheels that steer the work rather than gate it, turning “it feels better” into proof you can act on.
• Partner with and influence cross-functional teams to shape the technical roadmap, and drive issues to root cause when quality or operations are on the line.
• Raise the engineering bar with better patterns, better practices, and a standard other engineers want to match.