Senior Applied Scientists and Principal Applied Scientists (multiple Positions) - Copilot Tuning

Microsoft Microsoft · Big Tech · Redmond, WA +1 · Applied Sciences

Seeking Senior/Principal Applied Scientists to fine-tune LLMs on tenant data for M365 Copilot, creating task-specific agents and solutions. Role involves writing training pipelines, designing experiments, implementing inference solutions, and shipping models to customers. Focus on advancing LLM capabilities in an enterprise context.

What you'd actually do

  1. Write and execute training pipelines for large language models post-training.
  2. Design experiments to show the effectiveness of LLM-based solutions.
  3. Design and implement inference solutions that incorporate post-trained models following product specifications and work with broader team to ship these solutions to customers.
  4. Document experiments and communicate results across the team.
  5. Mentor early in career team members.

Skills

Required

  • Bachelor's Degree in Statistics, Econometrics, Computer Science, Electrical or Computer Engineering, or related field AND 4+ years related experience OR Master's Degree AND 3+ years related experience OR Doctorate AND 1+ year(s) related experience OR equivalent experience.
  • Ability to meet Microsoft, customer and/or government security screening requirements

Nice to have

  • 2+ years of experience training/fine tuning AI/ML models, preferably LLMs/SLMs (small learning model).
  • 2+ years of experience with Python and/or ML frameworks such as PyTorch.
  • 4+ years experience creating publications (e.g., patents, libraries, peer-reviewed academic papers).
  • 2+ years experience presenting at conferences or other events in the outside research/industry community as an invited speaker.
  • 2+ years of experience building Generative AI pipelines, e.g. with RAG (Retrieval augmented generation).

What the JD emphasized

  • training pipelines for large language models post-training
  • shipping high-quality models
  • customer data
  • generalize the learnings for the broader product
  • training/fine tuning AI/ML models, preferably LLMs/SLMs
  • creating publications
  • building Generative AI pipelines, e.g. with RAG (Retrieval augmented generation)

Other signals

  • fine-tune large language models (LLMs) on tenant data
  • task-specific agents and solutions
  • advancing the state of the art of models in M365 Copilot
  • shipping high-quality models
  • generalize the learnings for the broader product