Senior AI Researcher: LLM post-training for hallucination reduction (f/m/d)
Heidelberg • FullTime
Posted 3d ago
Remote Work Policy
On-site
Employment Type
FullTime
Categories
AI Research Engineer
About the job
Aleph Alpha is seeking a Senior AI Researcher to join our hallucination reduction capability team. This role focuses on pushing the boundaries of LLM post-training, specifically Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), to reduce model hallucinations and improve grounding in evidence. You will be part of a team conducting cutting-edge research on foundational models, with access to hundreds of GPUs, contributing to the development of trustworthy GenAI for industrial applications.
Responsibilities
- Design, execute, and scale post-training methodologies (SFT, RL) for hallucination elimination and evidence grounding.
- Research and implement techniques for uncertainty estimation, calibration, self-consistency, and self-correction in models.
- Build and optimize distributed fine-tuning and training pipelines for multi-node GPU clusters.
- Design and implement rigorous evaluation methods to benchmark hallucination rates, truthfulness, and reasoning accuracy.
- Independently drive research projects from hypothesis to scalable production implementation and technical reports.
Requirements
- Hands-on experience fine-tuning or post-training LLMs using modern distributed training frameworks.
- Research or implementation experience in hallucination mitigation, truthfulness alignment, grounding in RAG, uncertainty estimation, out-of-distribution detection, Retrieval-Augmented Generation evaluation/alignment, refusal behavior calibration, or claim verification.
- Strong programming skills in Python and PyTorch.
Benefits
- 30 days of paid vacation
- Access to fitness & wellness offerings via Wellhub
- Mental health support through nilo.health
- Substantially subsidized company pension plan
- Subsidized Germany-wide transportation ticket
- Budget for additional technical equipment
- Flexible working hours
- Hybrid working model
- JobRad® Bike Lease