Research Engineer Internship – Tool Benchmarking for LLMs

About the role

This research internship, conducted jointly with the MICS laboratory at CentraleSupélec, focuses on benchmarking and optimizing how language models use external tools. You will split your time between Alpic’s Paris office and the university lab, contributing to cutting‑edge AI‑native infrastructure.

Key responsibilities

• Investigate parameters that affect tool success rates in a black‑box LLM + MCP server environment.
• Analyze the impact of prompts, context windows, tool descriptions, error messages, model types, and server architectures on execution success.
• Propose and prototype new MCP server architectures, potentially incorporating reinforcement learning or fine‑tuning loops.
• Validate proposed solutions on Alpic’s production infrastructure and report results.

Required profile

• Currently enrolled in a Computer Science, Engineering, or related degree program at a tier‑1 university or engineering school.
• Creative and pragmatic mindset with a passion for shipping innovative AI products.
• Strong analytical abilities and interest in research‑driven development.

Required skills

• Experience with Large Language Models (LLMs) and prompt engineering.
• Knowledge of Reinforcement Learning concepts and fine‑tuning techniques.
• Understanding of MCP server architectures and AI‑native workflows.

What we offer

• Hands‑on research experience in a fast‑moving AI startup.
• Collaboration with leading researchers at CentraleSupélec.
• Opportunity to influence the next generation of AI‑driven internet infrastructure.