Bibek Gupta bibekgupta3333

👨‍💻 Bibek Gupta

Senior Software Engineer · Backend, Infrastructure, Cloud & Distributed Systems · AI/LLM Applications

🚀 About Me

const bibek = {
    role: "Senior Software Engineer",
    experience: "5+ years (since 2018)",
    companies: ["SLB Schlumberger", "Tekkon Technology", "Core Software Integrated", "FiberKnot Digital", "TechLekh"],
    industries: ["Fintech", "Agritech", "E-commerce SaaS", "Scientific Computing", "Subsurface Software"],
    focus: ["Event-Driven Microservices", "AWS Serverless", "Distributed Systems", "LLM Agents"],
    education: "MS Computer Science, Florida Polytechnic University ( May 2027)",
    research: "Published at ECSA-13, 2026",
    speaker: "AWS serverless architecture talk to 100+ attendees",
    openSource: true
};

What I build:

🔄 Event-driven and CQRS microservices for fintech: 40+ services handling about 500K events a day
☁️ AWS serverless and infrastructure as code (Lambda, DynamoDB, API Gateway, AppSync, CDK); led a modernization that cut monthly cloud cost by about 67%
🔧 High-concurrency backends (Node.js, TypeScript, Python): 75K+ users and 15K+ concurrent connections
🧪 Full-stack scientific tools: the CO2 Fast Simulator at SLB, with WebSocket-streamed progress and 55+ automated tests
🤖 Multi-agent LLM systems with RAG, evaluated against measured baselines

📈 Highlights


💸 ~95% fewer integration errors	Built event replay for a payment aggregate handling 10K+ daily fintech transactions
💰 ~67% lower AWS cost	Moved infrastructure to AWS CDK and modernized the Node.js runtime (about $15K → $5K a month)
🚨 ~40% faster MTTR	Improved observability and incident response on production microservices
⚡ ~45% faster APIs	PostgreSQL query optimization and caching on high-concurrency services
🎓 Graduate Teaching Assistant	Design and Analysis of Algorithms, Florida Polytechnic University

🧩 Featured Projects

Project	What it is	Results
FinSight AI	Multi-agent fraud detection that benchmarks six LLM agent-coordination patterns	91.2% F1 on the 6.36M-transaction PaySim dataset (vs 87.3% for a single-agent ReAct baseline); 1,150 tx/min at 3.12 s p95
Urban Event-Chain Prediction	Four domain agents plus a coordinator that predict seven-day event cascades across Chicago, NYC and San Francisco	621 labeled event chains · 3,825 predictions over a nine-model grid · seven baselines · paper at ECSA-13, 2026
env-manager	Cross-language development environment manager: a Python CLI and daemon with a React and WebSocket dashboard	Three operating systems · five language ecosystems · 15 pluggable tool adapters · shipped to PyPI and Homebrew
Code-Smell Study	Line-level code-smell detection with a large code model and dense RAG	0.964 micro-F1 · 473 expert labels across 21 smell types and four languages
CO2 Fast Simulator	Reservoir simulation SaaS for CO2 injection, with async jobs and P10/P50/P90 uncertainty analysis	Three unit systems · 55+ tests · deployed
PolyTrack (team)	Real-time microtransit telemetry: MQTT → FastAPI → WebSocket over PostGIS	Sub-second vehicle-position updates · deployed on AWS