13.10.2026 aktualisiert
100 % verfügbar
ML/LLM Engineer
Berlin, Deutschland
Nur Remote
Master's Degree in Logic and Computation, AI - TU ViennaSkills
Post-Training & Alignment
Expertise in RLHF, DPO, combined with LoRA, QLoRA, and PEFT methods. Hands-on experience with synthetic data generation and preference optimization for large language models up to 70B parameters.
Production Deployment of LLMs
Proficiency in deploying large language models using vLLM and TGI frameworks. Experience with model quantization, A/B testing, and regression tracking to ensure reliable and scalable production systems serving 50k+ users.
Safety & Evaluation
Strong background in adversarial testing, constitutional AI and boundary modeling, as well as evaluation frameworks. Skilled in designing comprehensive safety test suites and tracking behavioral consistency dimensions.
ML Infrastructure
Solid experience with PyTorch and HuggingFace Transformers for model development. Proficiency in Weights & Biases for experiment tracking, the Axolotl fine-tuning framework, and cloud platforms AWS and GCP.
Retrieval & Memory Systems
Knowledge in RAG pipeline design and vector search using FAISS and Pinecone. Experience with LangChain and vector embeddings for building retrieval-augmented generation systems.
Programming Languages
Proficiency in Python as the primary language for ML engineering, complemented by skills in C++, Java, and JavaScript for full-stack and systems-level development.
Full-Stack AI Integration
End-to-end LLM integration experience spanning Django backend APIs, prompt engineering, and React Native mobile clients. Skilled in building multi-agent systems and context management strategies for long-form generation.
Erfahrung
Berufserfahrung5 - 10 Jahre
Freelancer seit2024
Sprachen
DeutschMutterspracheEnglischverhandlungssicher
Projekthistorie
Data engineering and safety evaluation for behavioral alignment in dialogue systems. Designed synthetic data generation pipelines, executed ORPO and LoRA fine-tuning on 27B parameter models, modeled refusal boundaries, created adversarial attacker personas, conducted dataset ablations, implemented LLM-as-judge evaluation frameworks, and supported vLLM deployment to production serving 50k+ users.
Production LLM and multimodal AI systems for consumer mobile applications. Designed and deployed production inference pipelines for open-source LLMs, built multi-agent narrative generation systems, integrated IP-Adapter conditioning into image generation pipelines, owned end-to-end LLM integration with Django and React Native, and implemented context management strategies for long-form story generation.
Production ML systems for collaborative SaaS software. Led end-to-end development of ML-powered product features, designed and fine-tuned transformer-based models for text generation and grammatical error correction, built training and deployment pipelines, conducted large-scale feature engineering, and owned the complete model lifecycle including experimentation, validation, A/B testing, and production rollout. Notable contributions include a Graph Neural Network for B2B Upsell Prediction (2023) and a Grammatical Error Correction system (2022).