About Us
experial builds digital twins of real customers — so companies get instant feedback without waiting weeks for
surveys. Our clients include major German B2C brands in finance and retail. Backed by top-tier VCs, we
move fast. We're looking for an ML Research Assistant who wants to build things that work and last — from research to
production.
Your Role
Work closely with our ML team on applied research and development. The focus is on building durable
frameworks, not one-off experiments.
• VLMs & multimodal pipelines — evaluate, improve, deploy
• RAG systems — design, iterate, measure
• Evaluation frameworks — robust, repeatable methods for ML and agentic systems
• Research → production — turning experiments into maintainable components
AI agents are part of our toolkit — to speed things up, not replace the thinking.
What We're Looking For
Must Have
• Strong Python (OOP, LLMs/VLMs in code) and PyTorch
• Hands-on experience training or evaluating deep learning models
• Solid grasp of transformer architectures (NLP and/or CV)
• Comfort with CLI, git, SSH, and reproducible workflows
• Clear English communication
Ideal — or Willing to Learn Fast
• Both research and engineering experience — shipped things, not just papers
• RAG pipelines, evaluation frameworks, fine-tuning LLMs/VLMs/SLMs
• Agentic frameworks: LangChain/LangGraph, AutoGen, DSPy
• Evaluation & observability: Langfuse, MLFlow/WandB
• GPU environments (Slurm, AWS, serverless)
Fast learners who are honest about gaps are very welcome.
Work Mode & Compensation
• Fully remote
• Europe/Germany: Part-time — Minijob 5–10h/week (~€556/month tax-free) or HiWi 15–20h/week
(€14–€20/h)
• MSc thesis co-supervision possible if topics align
What We Offer
• Mentorship from ML engineers and product leads
• GPU infrastructure, real production problems
• Early co-authorship on research publications
• Path to joining the core team
How to Apply
Good fit > perfect fit. Send a CV (highlight projects and ML experience) and a short note on what you've built
and what you want to learn next.
■ < correo electrónico eliminado por razones de seguridad >