Machine Learning Engineer, Platform
RemoteUnited States or New York City, New York, United States
Job Summary
Turn pod needs into platform capabilities by identifying general patterns within urgent requests without premature abstraction. Own capabilities end-to-end, assuming full responsibility for behavior in production across every deployment that utilizes the feature. Engineer for production reality by navigating accuracy, latency, cost, and reliability across complex institutional environments. Raise the bar across the company by converting learnings from one team into inheritable platform assets. Serve customers on both sides of the wall, acting as true customers for project pods and domain experts for the workflows they power. Work at the research frontier with production stakes, transforming LLMs, RL fine-tuning, and agentic systems into capabilities that dozens of institutional workflows depend on simultaneously.
Required Qualifications
- You understand how machine learning actually works — not just the tooling, but the philosophy underneath: what a loss function really optimizes, how generalization breaks under distribution shift, why evaluation is where systems quietly go wrong.
- You live at the bleeding edge of modern AI, with hard-won instincts for squeezing the most out of LLMs and agentic systems — prompting, fine-tuning, tool use, and reasoning.
- You know when a fine-tuned segmentation model beats a VLM, when a rule engine beats both, and how to compose all three into a system more accurate than any single model.
- You treat frontier models as components to be measured, pushed, and engineered — never as magic.
- You have the platform instinct: you spot the general capability hiding inside three teams' specific requests — and know when generalizing is premature.
- You treat internal teams as real customers with real deadlines, and measure your success in their velocity.
- You're energized by building things that have never existed, and comfortable when the problem, the data, and the definition of success all have to be invented at once.
- You'll build the models behind both loops: Institutional Intelligence that compounds and A foundation model for construction documents.
- You'll build the machinery that keeps the promise of Continuous improvement, engineered: capturing production corrections, triaging failures to the component that caused them, and turning that signal into retraining and safe redeployment — automatically, across every use case.
- Turn pod needs into platform capabilities — find the general capability inside one team's specific, urgent request, without over-abstracting before the pattern is proven.
- Own capabilities end-to-end. There is no handoff: whoever builds the capability owns its behavior in production, across every deployment that uses it.
- Work at the research frontier with production stakes, turning LLMs, RL fine-tuning, and agentic systems into capabilities that dozens of institutional workflows depend on at once.
- Engineer for production reality, navigating accuracy, latency, cost, and reliability across environments far messier than any benchmark.
- Raise the bar across the company. The platform is how learnings travel: what one pod discovers, you turn into something every pod inherits.
Hiring someone like this?
Get your role in front of qualified candidates on Sorce.