I work on the layer between a model and the metal it runs on.
Computer Research Scientist, AI at Universitätsklinikum Tübingen. Five years of getting models to run somewhere real — distributed across a Slurm cluster, served through a traced inference pipeline, or running entirely offline on someone's own machine.
Focus — Distributed training on Slurm · LLM PEFT fine-tuning · Quantization & inference optimization · Agentic RAG
📍 Reutlingen, Germany · 🌐 orialpha.github.io · 💼 LinkedIn
| Autotrainer | Hand it a model and data — it finds the hardware, picks the distribution strategy, and infers the training recipe. PyTorch DDP, Slurm multi-node, TensorFlow, scikit-learn through one API. | Python |
| SlurmGenie | An offline copilot for Slurm GPU clusters. Diagnoses failed jobs, watches GPU utilization, rewrites sbatch scripts. Installs air-gapped. | Python |
| InfraSight | A transparent proxy that watches LLM, RAG and agent traffic — request logs, PII masking, nested agent traces, LLM-as-a-judge scoring. | JavaScript |
| The Vault | Local RAG that never leaves the machine. Ollama for generation, FAISS for millisecond retrieval, two models racing side by side. | Python |
| PivotDesk | Live pivot-point dashboard for NSE stocks, with a swing panel of moving averages, RSI, MACD, Supertrend and ATR. | Python |
Where I've worked
| Role | Where | When |
|---|---|---|
| Computer Research Scientist, AI | Universitätsklinikum Tübingen | Aug 2024 — present |
| Deep Learning Inference Engineer | Ella Lab GmbH, Köln | Apr 2022 — Jul 2024 |
| Machine Learning Engineer, Innovation | Clinomic GmbH, Aachen | Jan 2021 — Mar 2022 |
| Working Student, Data Science & AI | Aptiv Services Germany GmbH | Apr 2019 — Dec 2020 |
M.Sc. Embedded Systems, TU Chemnitz · B.E. Electronics & Communication, VTU
- LLM & retrieval — HuggingFace · LangChain · Ollama · FAISS · vector databases
- Training & inference — PyTorch · DDP · FSDP · TensorFlow · Keras · ONNX · TVM · scikit-learn · OpenCV
- Clusters & MLOps — Slurm · Docker · Kubernetes · Podman · Argo Workflows · GitLab CI · GitHub Actions · GCP · AWS
- Languages — Python · Bash · SQL · C
Open to ML infrastructure & engineering opportunities. Happy to discuss LLM fine-tuning, agentic pipelines, distributed training, cluster scheduling, or inference optimization.