snehit vaddi

Shipped a clinical ambient AI scribe serving 15,000+ providers across 11 specialties, automating 70% of documentation. Built agentic document processing with OpenAI Agents SDK and fine-tuned vision-language models, reducing costs from $400K/month to $20K/month (95%) while supporting PHI-safe pipelines and RAG quality guardrails.

Developed a hybrid YOLOv8-ViT model improving small-object detection by 15%, with Grad-CAM/EigenCAM explainability and publications in SPIE (2025) and IEEE (2023). Built an interactive React analytics dashboard (adoption 15% → 85%) and automated retraining via MLflow and GitHub Actions, reducing deployment time from 4 hours to 15 minutes.

Built a LangGraph-based multi-agent RAG system that autonomously resolved 65% of customer tickets from a 50K-doc knowledge base. Improved helpfulness from 43% to 76%, relevance by 30%, and achieved 92% recall@10 with FAISS hybrid search and vLLM (40% p95 latency reduction).

Engineered a BERT + XGBoost intent classifier (88% F1) that reduced unnecessary technician dispatches by 12K/year, saving $2M annually. Built Elasticsearch + Word2Vec anomaly detection to cut diagnosis time by 40% and Tier-2 escalations by 25%, and optimized PySpark/Delta Lake pipelines for 1M+ logs/day (30% lower latency).

Master of Science in Computer & Information Science from the University of Florida. Completed graduate research and published work in AI/ML and computer vision.

Grade: 3.9/4.0

Bachelor of Technology in Computer Science at GITAM University, completed with a GPA of 3.9/4.0.