Senior Machine Learning Engineer
Protege AIAbout the Role
Our next Senior Machine Learning Engineer will work on architecting the next phase of our core ML platform, the developer experience of creating models, and dogfood building models on our platform with clients. This is a hands-on tech-lead role where you’ll get to meet with customers and design their model pipeline from design docs and ideation to implementation and serving traffic.
Key Responsibilities
ML Model Development:
Experiment, develop, and create finetuned models that specialize in specific enterprise use cases for clients. For example, taking a generic customer workflow that uses SaaS tools and mapping it into an agent. The work will blend LLM post-training (SFT and RL-based optimizations), identifying the right harness, embedding retrieval, and similar to run the end to end for the client. Finally, handing the pipeline to the client for long-term ownership.
Evaluation & Benchmarking:
Build the evaluation harnesses and benchmarks that determine when a specialized model is production-ready. Balance the customization for end use cases with platform level features that are usable across teams.
Technical Leadership & Strategy:
Design the end-to-end ML Platform that powers development and optimizes the platform to maximize growth of models trained and hosted with us. Help drive the developer experience for customers and product roadmap with quarterly planning. Lead thorough and fair design doc reviews for new features.
Platform Excellence:
Monitor production systems built on our Bun/TypeScript backend and Postgres, troubleshoot issues, and ensure high availability SLAs through effective debugging and on-call rotations.
Collaboration & Innovation:
Translate product requirements into scalable ML designs by collaborating with product managers, full stack engineers, and UX designers. Stay ahead of trends like RL post-training, state of the art in open source, and sandbox platforms to keep our stack cost effective.
Continuous Improvement:
The ML stack moves quickly, and our products evolve with it. Knowing what to build and how to build it requires raising the bar of the entire team. You’ll help set a culture of continuous learning by providing fast feedback to individual team members, and encouraging experimentation within the team. Individually we expect you’ll regularly be exploring the latest industry trends and the means of how to stay up to date, from conferences to podcasts, NeurIPS papers, and endless scrolling on X.
Nice-to-Have Skills
(Highly Preferred) Familiarity with reinforcement learning for post-training.
Design and creation of coded reward verifiers for a variety of model use cases and class sizes
Development of simulators for reward calculation in single or multi turn environments
Familiarity with inference using quantized model artifacts; single endpoint multi-model serving directly on GPUs
Familiarity with trading off build vs buy managed services, like Together AI or Fireworks.
Qualifications
2+ years of experience of building models. Including manipulating raw data and workflow examples from end-specific applications into formatted training sets, then fine-tuning LLMs with sensible benchmarks to compare improvement against base and proprietary models.
Background in VC-backed startups or scaling ML solutions at large tech companies. (0→1) experience
Expertise in identifying appropriate agent harnesses, model candidates, and integrating numerous ML solutions for specific workflows.
5+ years of professional experience in software development, including a portion in tech lead roles