Machine Learning Engineer

Responsibilities

• Own the models inside Kepler's AI research platform: select which model runs each task, decide when a fine-tuned model beats a frontier one, and manage the training, evaluation, and extraction systems.
• Fine-tune small models on high-volume extraction tasks (e.g., footnote tables in 10-Ks, IR decks) and demonstrate improvements in accuracy, cost, and latency.
• Build evaluation harnesses that score agent research runs end-to-end (ensuring every number traces and every citation resolves) and integrate into CI to catch regressions.
• Redesign model routing across workflows: use frontier models for hard reasoning, cheaper or fine-tuned models for high-volume extraction and verification, backed by evals.
• Systematically improve workflows that succeed 80% of the time by identifying and addressing the remaining 20% through better tools, tighter verification rules, different context, fine-tunes, or model changes.
• Own systems end-to-end, from extending the platform to new industries to leading new architecture as infrastructure scales.
• Ship production systems with a focus on correctness, handling failure modes, regressions, subtle bugs, and debugging before demos.

Requirements

• 5+ years building production software (no upper limit; compensation scales with experience).
• Shipped ML systems to production, including fine-tuning, agents, retrieval, and structured extraction; understand what breaks between a demo and a product.
• Treat evals as engineering: build measurement before the feature and only call something better when numbers confirm it.
• Strong general engineering fundamentals, regardless of path into ML (research, ML infra, or product).
• Comfortable moving between a fine-tuning run and orchestrator code in the same day; able to work in a codebase you didn't write.
• Care about what analysts do with what you ship, not just whether the code was clever.
• Prefer fixing issues over filing tickets; proactively communicate design flaws before PRs.
• Strong communicator who anticipates problems and supports teammates without being asked twice.
• Thrive in a fast-paced environment where plans change frequently but work still ships.
• Low ego, willing to handle unglamorous problems and roll up sleeves in a small team.

Nice-to-Haves

• Experience with Rust (backend is Rust, but not required; strong fundamentals in other languages suffice).
• Background in finance, high-stakes industries, or building systems at scale (e.g., Palantir, Meta).

Compensation and Benefits

• Competitive compensation scaling with experience.
• 100% covered top-of-the-line medical benefits.
• Direct mentorship from engineers who built Palantir's core systems, including weekly 1:1s, architectural reviews, and a clear path to technical leadership.