Research Engineer — Scalable RL for Enterprise AI

Cerebras; San Jose (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Tessera Labs seeks a Research Engineer to own the model‑driven training, environments, and evaluation machinery at scale. You will build post-training stacks, memory and context systems, and knowledge graphs over real enterprise data, guiding experiments from concept to production.

You will work with Research Scientists to run end‑to‑end experiments, optimize throughput, and ensure reproducibility of results in a fast‑moving AI platform environment.

Qualifications

• Significant experience training, fine-tuning, or post-training language models.
• RL tuning experience (RLHF, RLAIF, RLVR, GRPO‑family, or agentic RL) is close to a requirement.
• Memory and context for long-running agents: architecture, retrieval, or training.
• Strong software engineering fundamentals; code for experiments must be runnable by others.
• Fluency in Python and PyTorch (or JAX); comfortable debugging distributed training.
• Design, run, and interpret experiments with empirical rigor; distinguish signal from noise.
• Experience with GPU infrastructure at scale; understanding time and memory usage.
• Want research to end up in production; treat that production constraint as a feature, not a bug.
• Clear written communication; decisions are documented.

Responsibilities

• Build and scale the post-training stack: SFT, preference optimization, and RL for long-horizon tool use and transformation over enterprise systems.
• Build memory and context machinery for long-horizon agents: retention, structure, retrieval and revision.
• Construct representation layers: ontologies and knowledge graphs from enterprise data and related pipelines.
• Design and implement data generation and curation pipelines: synthetic landscapes, traces, tool-call trajectories, curricula.
• Develop RL environments: sandboxed landscapes and verification harnesses with automatic scoring and partner traces.
• Create offline eval harness for long-horizon agentic behavior: trajectory scoring and reproducibility across weeks.
• Optimize training and inference throughput: kernels, parallelism, memory, batching, serving.
• Take training results from evaluation to serving traffic: quantization and rollback paths.
• Run experiments end-to-end: design, launch, debug, and analyze with emphasis on real effects vs. noise.

Job description

Tessera Labs seeks a Research Engineer to own the model‑driven training, environments, and evaluation machinery at scale. You will build post-training stacks, memory and context systems, and knowledge graphs over real enterprise data, guiding experiments from concept to production.

You will work with Research Scientists to run end‑to‑end experiments, optimize throughput, and ensure reproducibility of results in a fast‑moving AI platform environment.