Reddy Gurunatha Swamy Parasaram
Goes by Gurunath · AI Engineer
guruparasar [at] gmail [dot] com · LinkedIn · GitHub · Google Scholar
About
At Tiger Analytics I lead technical development of the agentic AI core of a multi-team healthcare payment-integrity program. It cut contract review from about an hour to seconds, made every contract in the portfolio searchable, and helped the program identify millions of dollars in overpayments. I own the AI system end to end: multi-agent design, LLM evaluation, AI guardrails for sensitive data, cost and latency, and MLOps/LLMOps in the cloud. I also lead client discovery and solution design for healthcare AI projects, work with business and legal leads on rollout and results, and coach teams on agentic coding and AI best practices.
This page is a short overview. The PDF has the full resume.
Experience
-
Tiger Analytics — Data Scientist
Built the AI core of a multi-team payment-integrity program that identified millions of dollars in overpayments, with millions recovered in year one, scaling automated audits from hundreds to thousands of hospitals. Lead the agentic extraction and retrieval system end to end: OCR, multi-agent LLM workflows with structured outputs and guardrails, PySpark pipelines on Databricks, an evaluation harness that cut evaluation from hours to 1 minute, and latency and cost work across the inference stack.
Stack: Python, SQL, PySpark, Databricks, LLM agents, OpenAI API, MLflow, Azure AI Foundry
-
EXL — Lead Assistant Manager (Senior Data Scientist)
Built hybrid RAG retrieval on GCP for an assistant used by thousands of service reps, raising search accuracy 60% and recall 50% through embedding selection and query augmentation. Fine-tuned intent and entity models and built offline evaluation pipelines.
Stack: Python, PyTorch, Hugging Face, PEFT, GCP, vector and hybrid search
-
University of Utah — Research Intern
Built tools for analyzing text embeddings in language models and the patterns in their high-dimensional representations.
Stack: Python, PyTorch, Hugging Face
-
Zoho Corporation — Member Technical Staff (Lead ML Engineer)
Led a team building a Transformer-based grammatical error correction system that processes tens to hundreds of millions of sentences per day. Scaled real-time inference to tens of thousands of requests per second with quantization, distillation and a CPU-to-GPU migration (50% lower latency, 7x throughput).
Stack: PyTorch Lightning, Transformers, ONNX Runtime, Triton, NVIDIA Nsight
Skills
- AI & LLMs
- LLM agents, multi-agent orchestration, RAG, structured outputs, guardrails, LLM evaluation, OpenAI API, LangChain, LangGraph, MCP
- Machine learning
- PyTorch, Hugging Face Transformers, PEFT, scikit-learn, quantization and distillation, ONNX Runtime, Triton
- Data analysis
- EDA, statistical analysis, slice and cohort analysis, SQL, PySpark, pandas, NumPy, Matplotlib
- Platforms
- Databricks, Azure, GCP, MLflow, Docker, FastAPI, GitHub Actions
- Healthcare AI
- HIPAA, PHI de-identification, RBAC, payment integrity
Education
MS in Computing, Artificial Intelligence
B.Tech in Electronics and Communication Engineering
Projects
- SalASR — Saliency methods for a fine-tuned Whisper speech recognition model.
- Explainable Deep Learning — Experiments with explanation methods for deep learning models.