Machine Learning Research Engineer

Every role at Polymodal is dual: founding platform builder first, with a ML research and signal intelligence edge that shapes the platform.

As agents take on more execution, the platform needs models that understand behavior, predict action, and improve from its own data. You'll build the signal intelligence that makes the platform adaptive.

Stack: Python/Rust, classifiers, eval pipelines, tiny predictive models, local / edge inference, inference routing, and world models for action prediction.

What you'll own

• Turn behavioral data into fast, tiny, useful models.
• Build and ship ML classifiers over agent and host telemetry.
• Build training/eval pipelines on our own data.
• Package models as production services (API, batch, edge).
• Optimize inference cost and latency: quantization, distillation, batching, caching.
• Partner with data engineering to turn raw signal into model-ready features.
  

About you

• You care about models that survive production, not just notebooks or benchmarks.
• You turn messy behavioral data into hypotheses, evals, and shipped systems.
• You move fast without losing scientific honesty.
• You work well where research, product, and systems blur.
  

Must-haves

• You use AI as part of your own research and engineering loop, not as a side tool.
• You have shipped ML into production and owned it end to end.
• Strong in Python/Rust and a modern ML framework (PyTorch or TensorFlow).
• Background in anomaly detection, sequence, or behavioral modeling.
• Experience with edge, embedded, or low-latency inference.
  

Nice-to-haves

• Experience with small-model training, distillation, or quantization.
• Publications, notable GitHub work, or hard side projects.
  

About Polymodal

Polymodal is a funded pre-seed company building the operating layer for agent-executed organizations. We closed our pre-seed in May 2026, backed by Lunar Ventures, Wind Capital, Volve Capital, and early-stage angels. We are early enough for every hire to shape the company, and real enough that the work must survive customers, scale, security, and trust.

Co-founders & team

You'll build directly with Joep (3x founder, cybersecurity), Sean (former CISO at Slack/Reddit/Bumble), Edo (2x founder in AR/XR and multi-agent systems), and Atul (founder, cybersecurity). We are future-facing, execution-focused, daring, and human.

Our offer

• Competitive salary and meaningful early equity, so you share real ownership.
• A chance to define a new category while product, architecture, and company are still malleable.