Artificial Intelligence

Designed an AI-enabled Security Operation Center (SOC) platform for real-time threat detection, User and Entity Behavior Analytics (UEBA), and playbook-driven remediation. Integrated Google Gemini AI for intelligent triage and built a live dashboard for monitoring.

Architected an AI-powered eligibility agent automating Medicaid eligibility assessments using LLM-assisted workflows, OCR document ingestion, and explainable AI decisioning across income, household size, disability, and residency attributes.

Designed and implemented a rule-based workflow to verify healthcare provider credentials prior to Medicaid enrollment. The workflow is automatically triggered automatically by the user intent.

Analyzed 264K+ FDA inspection records with XGBoost risk models and Groq LLM (Llama 3.3 70B) for natural-language querying. Built interactive dashboards with global heatmaps, facility search, and scenario-based compliance analytics.

Architected a machine-learning–based USDA Risk Prioritization Engine to predict operational and compliance risks across large-scale inspection and regulatory datasets. Built end-to-end AI/ML pipelines for feature engineering, risk scoring, and classification, with LLM-assisted extraction from unstructured inspection reports.

Built an LLM-powered agent generating ICD-10 and CPT code recommendations with confidence scoring and explainability. Reduced manual coding effort by ~40% through intelligent AI workflows with rule-based validation and clinical note OCR.

Designed and implemented an AI-powered OCR system to extract and structure prescription label information from medical images using Gemini LLM. Implemented automated parsing of medication details, patient information, pharmacy records, dosage instructions, and refill data.

Architected an AI-driven fraud detection platform leveraging machine learning models and feature engineering techniques to identify suspicious insurance claims. Trained predictive models using XGBoost and statistical learning methods to detect anomalous claim patterns and potential fraud indicators. Performed exploratory data analysis, feature engineering, and model validation to improve prediction accuracy and reduce false positives.

Architected and deployed a production-grade enterprise RAG system for intelligent knowledge management across large document repositories. Delivered hybrid search capabilities and context-aware LLM response generation with built-in hallucination guardrails and traceable source citations to power accurate decision-making.