sameer chachiya

▸ Building backend systems for cards & payments, focusing on risk evaluation and automated decision logic
▸ Designed risk scoring algorithms improving decision consistency in compliance-driven financial workflows
▸ Delivering low-latency, high-reliability systems under strict correctness constraints in a regulated environment

▸ Delivered AI-powered enhancements integrating LLMs (GPT, Claude) and agentic workflows into Java, Spring Boot, Python, and React —
enabling intelligent automation and contextual decision-making
▸ Designed and scaled 15+ microservices handling 100K+ daily requests at sub-50ms latency for production-grade AI inference
▸ Architected end-to-end RAG pipelines (embedding → vectorDB → retrieval → LLM inference) for context-aware production AI responses
▸ Led agentic AI adoption — agent-based workflows for autonomous task execution: code assistance, issue analysis, and workflow
orchestration
▸ Integrated LLM-driven automation into pipelines, increasing sprint throughput by ~3× and reducing debugging/triage effort by ~40%
▸ Improved query performance by 30% and UI responsiveness by 20% with zero latency regression on production systems
▸ Contributed to Angular-to-React migration as part of AI-ready architecture modernisation initiative
▸ Mentored engineers on AI integration strategies and microservice design, enabling team-wide GenAI capability uplift
▸ Hackathon Winner: Built GenAI + anomaly detection system, improving detection accuracy by 30%

▸ Developed 45+ secure applications in Java, Ruby on Rails, Spring Boot, and React JS across 5+ cross-functional global teams
▸ Created 10+ microservices integrating Kafka into an existing Monolith; containerised apps with Docker reducing cloud costs
▸ Automated software delivery with 15+ Java Groovy scripts achieving CI/CD, increasing delivery speed by 40%
▸ Developed 150+ unit test cases; participated in 100+ code reviews ensuring code quality and system stability
▸ Hackathon Winner: Automated CVE fixes, cutting engineering time from 14 hours to 30 minutes per CVE