I turn ambiguous problems into production systems — backend services, AI agents, and the pipelines that hold them together.
Results-driven Software Engineer with 3+ years of experience building scalable, production-grade systems across backend services and applied AI. Proven track record of measurable impact, strong ownership from design to deployment, and a mentoring mindset.
Java 21 + Spring Boot 3 HTTP service benchmarked at 267K RPS across a 3-JVM cluster, profiled with async-profiler flame graphs down to OS network I/O. Write-behind Redis cache (<1ms writes) with async PostgreSQL durability via virtual threads — 5x throughput over synchronous writes.
SHAP-guided adversarial defense for software vulnerability detection, published at ITNG 2025 (Springer). Cuts attack success rate from 86% to 0% while holding 85%+ clean accuracy.
Multi-agent loan evaluation platform with document processing, deterministic risk scoring, and generative-AI decision explanations.
Multi-agent RAG system with a 3-node LangGraph pipeline (retriever → answerer → critic) for independent verification. Dockerized FastAPI on AWS App Runner via ECR; ~850ms p50 latency with 8/10 regression eval pass rate.
100% serverless shoe marketplace on AWS — Lambda + API Gateway backend, DynamoDB for catalog, S3 for assets, Cognito auth, Stripe payments, CloudWatch observability, and GitHub Actions CI/CD.
AI-augmented marketplace for near-expiry goods with TensorFlow.js quality verification and dynamic pricing — pilot cut food waste 25% and lifted conversion 30%.
Academic portal streamlining course navigation and graduation planning for 9,000+ students, with a Selenium scraper against Ellucian Banner ERP.
Twitch-style VOD/Clips platform with HLS streaming, keyboard controls, clipping, and full keyboard accessibility.
Go + AWS serverless impression pipeline with idempotent composite-key writes, a DLQ, and full Terraform IaC.
SAAB Framework: a SHAP-guided test-time defense combining a BiLSTM + attention encoder, a Platt-calibrated linear SVM, and KernelSHAP attribution profiling to detect adversarial code edits without retraining the backbone model.
Open to backend, AI infrastructure, and full-stack engineering roles. My inbox is always open — say hello.