Research Scientist

Job Title: Research Scientist
Location: Seattle, WA (Remote)
Duration: 12 months contract with possible of extension

Description
This role is for a Research Scientist who will be responsible for designing and implementing robust processes to evaluate and improve LLM-based models. More specifically, they will work on ensuring that we build models that will accelerate the developer work cycle and improve how software is built.


This researcher will:
Work closely with engineers, PMs, designers, and data scientists to align on research goals and build scalable end-to-end evaluation pipelines.
Report on their research and make recommendations to stakeholders.
Work autonomously and mentor other researchers to support the goals of the research organization.

Minimum Qualifications
4+ years of experience in an applied research setting, or similar.
Proven experience developing benchmarks and metrics to evaluate AI models, preferably with a focus on code generation or natural language processing (NLP).
Familiarity with software engineering principles (unit tests, CI/CD, version control etc.).
Strong programming skills in 1 or more programming languages (including Python).
2 years or more of experience with machine learning and data science algorithms and tools (e.g., PyTorch).
Ability to produce meaningful visualizations and reports, and effectively present results to different audiences and across multiple media formats.
Ability to work independently as well as collaboratively with minimal direction.
Organized, highly attentive to detail, and manages time well.
Experience working in industry.

Preferred Qualifications
Advanced degree (MS/PhD) in data analytics, statistics, or any engineering or computer science field.
Experience with large language models (LLMs) and knowledge of state-of-the-art code generation techniques.
Publication(s) in the fields of artificial intelligence, machine learning or data science.
Familiarity with code generation-specific benchmarks and metrics.
Ability to modify or fine-tune pre-trained models for specialized tasks.
Familiarity with code scanning tools.