Sep 2025 – Feb 2026 · 5 mos

Jun 2023 – May 2024 · 11 mos

• Built a Retrieval-Augmented Generation (RAG) pipeline using Amazon S3 to store pre-chunked documents and Titan Embeddings. Instead of a dedicated vector DB, embeddings are stored alongside text in S3 and compared in-memory using cosine similarity within AWS Lambda, enabling a lightweight, cost-efficient retrieval system for resolving ServiceNow tickets.
• 📥 Webhook-based Ingestion: Integrated with ServiceNow via webhooks to capture new incident tickets in real time.
• 🧠 GenAI-Powered Resolution: Utilized Claude v2 for reasoning and response generation, applying a confidence threshold to ensure accuracy and fallback control.
• 🔍 RAG Pipeline: Built a custom pipeline using Amazon Titan Embeddings and chunked knowledge sources (KEDB, runbooks, logs), stored in Amazon S3, to retrieve top relevant context before response generation.
• ⚙️ Serverless Architecture: Orchestrated with AWS Lambda for scalable execution; integrated Amazon SQS for asynchronous queue handling and decoupled processing.
• 🔐 Secure by Design: Implemented AWS IAM roles and fine-grained policies for least-privilege access across services.
• 🚀 CI/CD & Deployment: Automated deployment and versioning using GitLab CI/CD pipelines, ensuring reproducibility and governance.

Jun 2022 – May 2023 · 11 mos

• Designed and deployed a location-based hospital recommendation system for Medibank, enabling support executives to suggest the best hospitals based on customer location, coverage eligibility, and service availability.
• 📍 Geo-intelligent ML Model: Implemented a K-Nearest Neighbors (KNN) classifier to identify the top nearby hospitals based on GPS coordinates and Medibank-approved provider lists.
• 🛡️ Insurance Matching Logic: Filtered results based on the customer’s active Medibank policy to ensure only in-network hospitals are recommended.
• 📈 Interaction Logging: Captured historical recommendation decisions and outcomes, stored in Amazon S3/RDS for continuous learning and auditability.
• 🌐 User-Facing Portal: Built using Django, providing support teams a responsive UI to input patient details and view AI-powered hospital suggestions.
• ☁️ AWS Architecture:
• Data Lake: Used Amazon S3 to store structured logs, user profiles, and hospital metadata.
• ML Pipeline: Deployed KNN model using AWS Lambda for lightweight, serverless predictions, with option to scale via Amazon SageMaker.
• APIs & Access: Managed APIs with API Gateway; enforced secure access via IAM roles.
• Monitoring: Included logging and tracing for model calls and UI activity to improve performance and trust.

Jun 2021 – Sep 2025 · 4 yrs 3 mos

• As a GenAI Engineer at TCS, I design and deploy AI-powered solutions for leading healthcare clients including Medibank, CVS Health, AbbVie, and Optum. My work spans:
• 🧾 ServiceNow Automation: FLR systems using Claude, Titan embeddings, and AWS Lambda
• 📍 Geo-ML Solutions: Hospital recommendation engines with KNN classifiers and insurance matching
• 🔁 RAG Pipelines: Titan-powered document retrieval using embeddings stored in Amazon S3, with inference on Lambda
• ☁️ AWS Stack: Lambda, S3, SQS, API Gateway, IAM, and GitLab CI/CD for secure, scalable deployments
• 🔐 Healthcare Focus: Solutions aligned with privacy, compliance, and performance in high-trust environments
• I specialize in building GenAI systems that enhance operational efficiency, streamline support workflows, and improve healthcare outcomes through intelligent automation.

Jun 2021 – May 2022 · 11 mos

• 🔧 Key Contributions:
• Developed Python scripts to extract structured data from scanned PA documents using AWS Textract, regex, and rule-based logic.
• Applied BERT-based models (e.g., BioBERT) to extract medical entities (diagnoses, procedures, medications) from clinical narratives in free text.
• Built a lightweight document similarity engine using cosine similarity over pre-computed BERT embeddings, enabling contextual lookups of past approvals stored in Amazon S3.
• Implemented business logic for coverage checks using patient plan data and MedDRA codes, enabling auto-triage for straightforward approvals/denials.
• Deployed components via AWS Lambda, orchestrated using SQS, and enforced secure role-based access via IAM.
• Contributed to GitLab CI/CD pipelines for automating code and model deployments.