Gururaj Somanna

I’m an ML Engineer and Data Scientist with 4 years of experience building production-grade AI systems that deliver measurable business impact. I specialize in LLM/RAG pipelines, clinical NLP, and end-to-end MLOps, with results that translate directly to performance and cost.

In healthcare, I architected production RAG using LangChain and Pinecone over 100k+ clinical documents, improving retrieval accuracy by 35%. I also fine-tuned BioBERT for medical entity and dosage extraction from unstructured physician notes, achieving a 94% F1-score and eliminating 500+ hours/month of manual data entry.

Across fraud prevention and data-intensive modeling, I built an autoencoder-based anomaly detection model on real-time Apache Kafka streams that prevented $50k+ in annual revenue leakage. I engineered 500+ features from multi-terabyte healthcare datasets with PySpark, reducing training time from days to hours, and I created a multilingual RoBERTa intent classifier reaching 91% accuracy and increasing automated ticket resolution by 40%.

I bridge research and production—using techniques like Vision Transformers for defect detection, synthetic data generation with Stable Diffusion + ControlNet, and real-time inference on edge devices. I’m hands-on with AWS (EC2, S3, Lambda), Docker, and CI/CD to ensure models are reliable, deployable, and maintainable in real environments.