Machine Learning Engineer (All Levels)
Location(s)
This is a hybrid role based out of our San Francisco, CA or New York, NY office. Our team works in-office time on Mondays, Wednesdays, and Fridays.
About the Role
Casap is hiring a Machine Learning Engineer to build and ship agentic systems, and to own how we measure whether they actually work. This is a backend-leaning ML role: you will write production code, not just prototypes.
We already have a strong foundation in classical ML, from model building and system design through to productionizing models. Our next chapter is agentic: building LLM-powered agents and creating the evaluation systems that let us trust them in production.
The highest-priority focuses of this role are evaluations: designing, building, and operating the infrastructure that tells us whether our LLM and agentic systems are getting better or worse, and rigor around distributed systems. You will work closely with the hiring manager and other engineers, with real ownership over both the agents and the systems that evaluate them.
What You'll Do
- Design and build evaluation systems for LLM and agentic workflows, including datasets, metrics, automated graders (including LLM-as-judge), and regression suites.
- Productionize those evaluation systems so they run reliably at scale, in CI and against live traffic, and give the team fast, trustworthy signals on every change.
- Build agentic systems end to end: tool use, planning, orchestration, memory and context management, and failure handling.
- Analyze agent behavior and failure modes, then turn what you find into better prompts, tools, architectures, and evals.
- Build and scale the backend and distributed systems (pipelines, queues, services) that agents and evals run on.
- Partner on our existing classical ML systems, helping take models from design to production and keeping them healthy.
- Raise the bar for engineering quality through code review, design docs, and sharing what you learn.
- Trajectory analysis and observability for LLM applications.
What You’ve Done
- Built and operated distributed systems or production backend services, and are comfortable reasoning about scale, reliability, and latency.
- Evaluated LLM or agentic systems in production: built eval harnesses, defined metrics, curated golden cases, and used the results to drive product decisions.
- Built and shipped LLM-based or agentic systems in production, and have a clear view of what makes them reliable or brittle.
- Trained and deployed classical ML models, with experience in problem framing, features, validation, and monitoring.
- Comfortable owning ambiguous problems from idea to production, and communicating clearly about tradeoffs.
Nice to Have
- Experience building evaluation infrastructure for agents specifically (trajectory-level evals, simulated environments, online/offline eval parity).
- Experience with workflow orchestration and large-scale data processing frameworks.
- Familiarity with human-in-the-loop labeling and annotation workflows for building eval sets.
- Open-source contributions or published work related to LLMs, agents, or evaluation.
About the Company
More jobs at casap
-
Engineering Manager, Integrations
New York, USA · hybrid · Nov 6, 2026
-
Staff Software Engineer
New York, USA · hybrid · Oct 27, 2026
-
Full Stack Engineer (All Levels)
New York, USA · hybrid · Oct 22, 2026
-
SRE - Database Reliability (All Levels)
New York, USA · hybrid · Oct 8, 2026
-
Backend Engineer (All Levels)
New York, USA · hybrid · Aug 19, 2026