Generative AI Engineer at JDIT Software Solutions Pvt. Ltd. (2025-01 – Present)
Project: Agentic AI & Clinical RAG Platform — Client: Philips
- Designed and deployed autonomous multi-agent workflows in Python using LangGraph and LangChain, automating a majority of unstructured clinical data processing and reducing the clinical team's manual review workload.
- Developed Python orchestration services and REST APIs (FastAPI, Boto3) connecting agentic workflows to Amazon Bedrock (Anthropic Claude), standardizing tool-calling across agents and accelerating development.
- Engineered automated Python data pipelines from Bedrock to Amazon QuickSight, replacing manual/batch reporting with near real-time clinical analytics dashboards.
- Architected a RAG pipeline over a large corpus of clinical records using Amazon Bedrock, improving retrieval relevance and end-to-end response latency.
- Configured Amazon Bedrock Guardrails and custom Python prompt-safety filters to block prompt injection and unsafe content with minimal added latency.
- Fine-tuned domain-specific open-source LLMs (7B/13B) using LoRA/QLoRA in Python, reducing GPU memory requirements and improving NER performance on evaluation data.
- Wrote Python scripts to curate, clean, and validate large training/evaluation datasets, contributing to fewer hallucinations observed in production.
- Partnered with cross-functional stakeholders to translate business requirements into GenAI solution design and delivery.
Software Engineer at LTIMindtree (2021-07 – 2025-01)
Project: CGMS (Continuous Glucose Monitoring System) — Client: Dexcom
- Designed and managed scalable, secure, highly available infrastructure on Google Cloud Platform (GCP), with IaC via Terraform and Deployment Manager supported by Python automation scripts.
- Built and optimized CI/CD pipelines using Cloud Build, GitHub Actions, and Jenkins, including custom Python build and deployment scripts.
- Implemented and managed Google Kubernetes Engine (GKE) clusters, including auto scaling, monitoring, and security configuration (IAM, VPC Service Controls, Cloud Armor, ITIL).
- Wrote Python-based Cloud Functions and monitoring scripts for proactive alerting in Cloud Monitoring, catching the majority of CPU/memory bottlenecks before they caused SLA breaches.
- Supported microservices architecture and service mesh using Cloud Run and Cloud Functions, with Python as the primary runtime language.
- Led production incident response and root cause analysis (RCA), building Python log-analysis tooling that improved MTTD and MTTR across GKE services.
Software Engineer at LTIMindtree (2021-07 – 2025-01)
Project: Anaplan Upload — Client: Johnson & Johnson
- Provided application support and monitored metrics/logs using Python scripts to proactively resolve performance bottlenecks within SLA.
- Analyzed performance metrics and user feedback, and managed tickets/service requests to drive continuous improvement.
