• Improve agent performance: Diagnose failure modes, refine routing/planning/tool-use strategies, optimize prompts, reduce latency without hurting quality.
• Own evaluation sets: Design balanced, adversarial, and scenario-based evals; build/extend an automated evaluation harness; track win-rates, pass@k, precision/recall/F1, and task success.
• Run applied AI research: Scout papers/techniques, reproduce bite-size results, prototype ideas (self-reflection, planning, retrieval augmentation, tool composition), and ship what works.
• Experiment & measure: A/B test agent changes, define guardrails, and land improvements with clear dashboards and acceptance criteria.

What You’ll Work With Python

• SQL
• Git
• RAG Pipelines
• Document Parsing/Extraction Pipelines
• Eval Pipelines
• Jupyter
• Experiment Tracking
• Prompt/Agent Tooling
• Retrieval & Tool Orchestration
• Metrics Dashboards

What We’re Looking For Must-haves: Analytical mindset, attention to detail, crisp communication, comfort with spreadsheets/Docs, and a bias for rapid iteration.

Nice-to-haves: Strong Python and SQL foundations; familiarity with evaluation metrics; curiosity about LLM planning, retrieval augmentation, and agent architectures.

Bonus signals: A small repo, AI research paper summary, or write-up where you improved an agent/model or designed a solid eval.

Why you’ll love it

• Impact from day one: Your changes roll into agents real customers use.
• Mentorship: Tight feedback loops with ML, Engineering, and Product.
• Ownership & speed: Small team, meaningful problems, quick ships.
• Hybrid flexibility: Deep-work time, remote, collaboration days, downtown.

Logistics & eligibility Language: English (French an asset).

Targeted fields of study: Engineering, Computer Science/IT, Physical Sciences.

How To Apply Deadline: November 15, 2025, 11:59 PM ET (applications reviewed on a rolling basis)

Submit: Resume + transcript + Cover Letter (optional)