AI Researcher: LLM post-training for hallucination reduction

Aleph Alpha’s mission is to deliver sovereign, human-centric AI to governmental institutions and large enterprises. To that end, we deliver category-defining AI innovation that enables open, accessible, and trustworthy deployment of GenAI in industrial applications. Our organization develops foundational models and next-generation methods that make it easy and affordable for Aleph Alpha’s customers to increase productivity in development, engineering, logistics, and manufacturing processes.

We are looking for a (Senior) AI Researcher to join our hallucination reduction capability team to push the boundaries in LLM post-training (SFT and RL).

We're one of the few places in Europe training foundation models from scratch. You will be part of a team of (senior) researchers and PhD interns doing cutting-edge research across pre-, mid- and post-training, grounding and specialized models with hundreds of GPUs at our disposal.

• Core Research & Development: Design, execute, and scale post-training methodologies (SFT, RL) specifically aimed at eliminating model hallucinations and making them grounded in evidence.
• Uncertainty & Reliability: Research and implement techniques in uncertainty estimation, calibration, self-consistency, and self-correction mechanisms to enable models to identify and signal when they “don't know”.
• Post-Training Pipelines: Build and optimize high-throughput, distributed fine-tuning and training pipelines across multi-node GPU clusters.
• Evaluation & Benchmarking: Design and implement rigorous evaluation methods to benchmark hallucination rates, truthfulness, and reasoning accuracy.
• Technical Ownership: Drive research projects independently from initial hypothesis to scalable production implementation and technical reports.

• Hands-on experience fine-tuning or post-training LLMs using modern distributed training frameworks.
• Track record of research or implementation in at least one of the following domains:
• Hallucination mitigation / truthfulness alignment / grounding in RAG
• Uncertainty estimation, or out-of-distribution detection
• Retrieval-Augmented Generation evaluation / alignment, refusal behavior calibration, or claim verification
• Strong programming skills (Python & PyTorch).

• Become part of an AI revolution!
• 30 days of paid vacation
• Access to a variety of fitness & wellness offerings via Wellhub
• Mental health support through nilo.health
• Substantially subsidized company pension plan for your future security
• Subsidized Germany-wide transportation ticket
• Budget for additional technical equipment
• Flexible working hours for better work-life balance and hybrid working model
• JobRad® Bike Lease