Full-stack engineer with 3.7 years across healthcare (Optum), fintech (BNP Paribas), and academic platforms (USF) — and a parallel track in applied AI research. I build production systems end-to-end: API design, cloud deployment, observability, and on-call ownership.
On the AI side, I'm more interested in correctness and explainability than benchmark scores. Every system I build includes citation grounding, confidence thresholds, or structured failure modes — because a system that fails loudly is more useful in production than one that fails silently.
Currently exploring: multi-agent coordination patterns, vLLM inference optimisation, Rust for systems programming.
Full stack breakdown
| Layer | Technologies |
|---|---|
| Languages | C#, Java, Python, TypeScript, JavaScript (ES6+), C++, PowerShell |
| Backend | ASP.NET Core, Spring Boot, FastAPI, Django, EF Core, Hibernate |
| Frontend | React, Angular, Next.js, Redux Toolkit, TanStack Query, Tailwind CSS |
| AI & ML | LangGraph, PyTorch, PEFT, LoRA, OpenAI GPT-4, Pinecone, HuggingFace, Ollama |
| Cloud | Azure AKS, Azure Functions, Cosmos DB, AWS (EC2/S3/Lambda/SQS), Docker, Kubernetes |
| Data | Apache Kafka, SQL Server, PostgreSQL, Redis, MySQL, Cosmos DB |
| Security | JWT, OAuth 2.0, RBAC, Spring Security, ASP.NET Identity, Azure Key Vault |
| Tools | Git, Postman, VS Code, IntelliJ IDEA, Jira, Linux/Bash, GitHub Actions |
| 🏥 Healthcare | 🎓 Academic | 🏦 Fintech | 🤖 AI Research |
|---|---|---|---|
| 3,500+ events/min Kafka | 5,000+ daily users | 30% fewer errors | F1: 0.589 ECG benchmark |
| 35% MTTR reduction | 820ms→310ms API | 20 hrs/week saved | 60% fewer parameters |
| 100K+ patient records | 99.6% uptime | 4 trading desks | 2.1× train efficiency |
🏥 Optum Kafka pipelines at 3,500+ events/min · MTTR reduced 35% · 100K+ patient records
🎓 USF 5,000+ daily users · 820ms→310ms API latency · 99.6% uptime · 60% faster page load
🏦 BNP Paribas 4 trading desks · 30% fewer reconciliation errors · 20 hrs/week saved
🧠 ECG-PEFT Wav2Vec2+LoRA: F1 0.589 / AUC 0.620 · 60% fewer trainable parameters
🤖 ResearchFlow Cited answers in <30s · Stateful LangGraph retry on confidence threshold
🔒 ContextFlow 100% on-device inference · Zero cloud API calls · Sub-second on consumer hardware
ResearchFlow AI — Agentic LLM Research Platform
Self-correcting LangGraph agent: Planner → Decomposer → Search → Verifier → Synthesizer. Delivers citation-grounded answers in under 30 seconds via stateful graphs with conditional retry edges on confidence-scored results — not possible in a linear chain.
Python LangGraph FastAPI Tavily React TypeScript Ollama
ECG-PEFT Bench — Medical AI Benchmarking
Benchmarked 4 PEFT strategies across 3 cardiac foundation models on 10K+ ECG segments. Wav2Vec2 + LoRA achieved best F1/AUC (0.589 / 0.620) with 60% fewer trainable parameters — structured model-selection report for clinical AI pipeline decisions.
Python PyTorch HuggingFace PEFT Wav2Vec2 HuBERT ECG-FM
DocuMind — RAG Document Intelligence
Pinecone-backed RAG platform with GPT-4 generation and citation grounding — every answer surfaces the exact source passage for auditable, verifiable multi-document Q&A with sub-second retrieval latency.
Python FastAPI Pinecone OpenAI GPT-4 React TypeScript
ContextFlow AI — Privacy-First Local LLM Chrome Extension
100% on-device inference via Ollama + Llama 3.2 — zero cloud API calls, no data leaves the browser. Manifest V3 service worker as persistent agent runtime with dynamic tool routing and sub-second response times.
TypeScript React Chrome Manifest V3 Ollama Llama 3.2 Vite Tailwind CSS
Workforce management for 500+ employees with optimistic locking for concurrent shift-assignment safety. GenDiff-PEFT — 2.1× training efficiency, 60% fewer parameters, FID within 8% of full fine-tuning. CLAP-AudioLDM — 15%+ CLAP alignment improvement via multi-candidate quality-aware selection.
| Company | Role | Period |
|---|---|---|
| University of South Florida | Application Developer — .NET · React · Azure | Aug 2024 – Present |
| Optum (UnitedHealth Group) | Associate Software Developer — .NET · Angular · Azure | Aug 2023 – Aug 2024 |
| BNP Paribas | Software Engineer — Spring Boot · React · AWS | Feb 2023 – Jul 2023 |
| Rinex Private Limited | Full Stack Intern — Spring Boot · React | Jun 2021 – Dec 2021 |
| Degree | Institution | GPA |
|---|---|---|
| M.S. Computer Science & Engineering | University of South Florida | 3.95 / 4.0 |
| B.Tech. Computer Science & Engineering | Amrita Vishwa Vidyapeetham | 8.31 / 10 |
Open to SWE · AI/ML Engineer · Full-Stack Engineer roles — US-wide, available June 2026
