Profile
Senior AI/ML Engineer and Data Scientist with 8+ years of experience designing predictive models, distributed data pipelines, and enterprise Generative AI architectures. Specialized in secure hierarchical agents, GraphRAG and multi-agent retrieval, multimodal document intelligence, explainable ML, and production LLM evaluation.
Experienced in regulated financial and healthcare environments, including PII/PHI controls, deterministic guardrails, transparent decisioning, high-availability Kubernetes deployments, and continuous semantic-drift monitoring.
Professional experience
Agentic Underwriting Intelligence & Borrower Servicing Automation
- Reduced manual underwriter document-review time by 43.2% by architecting a VLM pipeline for 1.4M+ unstructured financial documents annually.
- Delivered a 28.7% Tier-1 call deflection rate through GraphRAG retrieval across servicing guidelines and regulatory policies.
- Reached sub-920ms response times by integrating secure, parameterized MCP tool calls with enterprise loan-origination systems.
- Sustained zero downtime across 2.1M+ monthly inference requests using LangGraph, Docker, Kubernetes, and AWS EKS.
- Achieved a 99.99% zero-leakage audit score through streaming PII redaction and deterministic guardrails.
Core technologies: Python, LangGraph, GraphRAG, VLMs, MCP, LangSmith, FastAPI, AWS EKS, Kafka, Kubernetes
AI Engineer
Humana · Dallas, Texas
Agentic AI Platform & Member Experience Automation
- Deployed a LangGraph, Vertex AI, and Gemini hierarchical-agent platform to 18,400+ active member advocates.
- Improved Medicare eligibility retrieval accuracy by 42.7% and reached sub-850ms multi-stage latency using MRAG, E5 embeddings, and cross-encoder reranking.
- Reduced structured data-retrieval time by 28.4% through MCP-based access to PostgreSQL and EHR systems.
- Maintained 99.95% uptime across 1.2M+ daily AI API requests on GCP and Kubernetes.
- Reduced flagged inaccurate outputs by 31.5% using continuous RAGAS and LLM-as-a-Judge evaluation.
Core technologies: Python, LangGraph, Vertex AI, Gemini, MRAG, FastAPI, MCP, RAGAS, GCP, Kubernetes, PostgreSQL
Institutional Credit Scoring & Automated Underwriting Engine
- Achieved 99.88% validation accuracy for institutional credit-risk assessment with tuned XGBoost and LightGBM models.
- Implemented SHAP explanations that supported risk-audit approval with zero Q2 regulatory violations.
- Reduced financial reconciliation latency from 45.3 minutes to 9.2 minutes using Kafka and Snowflake.
- Served 12,400+ institutional participants with a sub-112ms P95 FastAPI inference endpoint.
Core technologies: Python, XGBoost, LightGBM, SHAP, BERT, Databricks, MLflow, Snowflake, Kafka, FastAPI, AWS
Enterprise Predictive Analytics & Big Data Platforms
- Reduced daily batch execution from 52 minutes to 14 minutes across 34.7M daily events with PySpark and Airflow.
- Reduced retail inventory stockout anomalies by 23.4% using Scikit-learn and Random Forest forecasting.
- Improved fraud true-positive detection by 19.2% with real-time anomaly classifiers.
- Cut multi-cloud deployment time by 35% using Docker and GitHub Actions across AWS SageMaker and Azure ML.
Core technologies: Python, PySpark, Scikit-learn, XGBoost, Airflow, AWS SageMaker, Azure ML, Docker, PostgreSQL, Redis
Associate Software Engineer — Data Science
Dozee · Bengaluru, India
Contactless Remote Patient Monitoring & Predictive Health Analytics
- Reduced environmental artifact noise by 31.4% through digital filtering of Ballistocardiography signals.
- Reached 94.6% classification accuracy for respiratory anomalies with 1D-CNN and LSTM architectures.
- Predicted acute clinical deterioration trends an average of 4.2 hours before visible symptom onset.
- Reduced non-critical alert fatigue by 18.7% across 5,640 monitored hospital beds.
- Sustained sub-240ms processing latency through Dockerized Flask services on AWS EC2.
Core technologies: Python, TensorFlow, Keras, LSTM, SciPy, Pandas, NumPy, Flask, AWS, Docker, signal processing