xelys jobs xelys jobs

AI Research & Data Evaluation Work Student (Part Time)

Lilt

full-remotejuniorpermanentdatabackend Full remote - Berlin, DE 190 days ago via WTTJ

See how well this job matches your profile

Sign up to get an AI match score and generate a tailored application in seconds.

Get your match score

Tags

Large Language ModelsLLM EvaluationBenchmarkingPythonNLPMachine LearningAutonomous AgentsExperimentationGitBash

About the role

Role Overview

Part-time AI Research & Data Evaluation Work Student (Full remote, Berlin). You will evaluate and benchmark frontier LLMs and autonomous agents, create/modify benchmark datasets, and design and run experiments—working closely with models and top-lab teams.

Key Missions

  • Stress-test advanced models to identify weaknesses and improve performance
  • Run rigorous evaluations and benchmarks on LLMs and autonomous agents
  • Create or modify benchmark data to probe reasoning and linguistic limits

Responsibilities

  • Evaluate and benchmark frontier LLMs and agents
  • Design and execute experiments
  • Build or update benchmark datasets and evaluation materials

Requirements

  • Currently enrolled at TU Berlin (Bachelor/Master) in Computer Science or related field
  • Strong drive to deliver customer-facing work with high quality, sometimes under tight deadlines
  • Quick learning mindset for new research methods and models
  • Highly proficient in Python, Bash, and git
  • Proficient in English
  • Solid understanding of LLMs, NLP, or machine learning

Nice to Have

  • Proficient in one or more non-English languages

Benefits (as stated)

  • Medical coverage, paid parental leave
  • Monthly lifestyle benefit stipend
  • Meaningful equity
  • 401(k) matching
  • Flexible time off

About Lilt

Lilt is an AI company focused on building and applying artificial intelligence, including work involving large language models and autonomous agents. The role indicates Lilt collaborates with leading AI research ecosystems and contributes to evaluation and benchmarking efforts for industry-standard models.

Scraped 5/20/2026