Machine Learning Engineer & Medical AI Researcher
Building high-performance, reproducible, and privacy-preserving multi-modal AI for real-world clinical impact.
I specialize in Multi-Modal AI, Whole Slide Image (WSI) Analysis, Digital Pathology, Radiology, and LLM/SLM Workflows. My focus is bridging the gap between cutting-edge AI research and scalable, edge-ready deployments.
- β‘ High-Performance Inference & Edge AI: Optimized deep learning models yielding up to 5.4x inference speedup and 80% memory reduction for low-resource CPU execution.
- π Multi-Modal Vision-Language Systems: Principal investigator/developer on Google Cloud TPU & NIH U24 grants building Vision-Language Models (VLMs) and privacy-preserving Small Language Models (SLMs) for clinical workflows.
- π¦ Open Source & Standardization: Key maintainer and core developer behind GaNDLF (MLCommons) and sauron.
- π§ Federated Learning & Benchmarking: Co-lead on international benchmarking efforts (e.g., FeTS, BraTS-Path).
Languages β Python, C++ Frameworks β PyTorch, PyTorch Lightning, Scikit-Learn, LangChain Deployment β OpenVINO, TensorRT, Docker, ONNX Domain AI β Vision-Language Models (VLM), SLMs/LLMs, 3D Medical Imaging, DICOM, WSI Tooling β Git, PySpark, High-Performance Compute (HPC / Cloud TPUs)
Science works best when grounded in humanity. Outside of research:
- Culinary Experiments: Cooking is just chemistry you can eat! π³
- Human-Centric Tech: Deep believer in empathetic engineering and discussions on why feelings matter more than facts when designing user-facing AI systems.
- π§ Email: thakursp@iu.edu / siddhesh0011@gmail.com
- π Location: Indianapolis, IN, USA
- π Google Scholar: Siddhesh Thakur