• Design and build REST APIs with clear contracts, layered service boundaries, and explicit error handling
• Apply correctness-first patterns — idempotency keys for deduplication, pessimistic locking for concurrent writes, and audit trails for observability
• Structure Spring Boot applications with separation of concerns: controller → service → repository, with profile-based config for dev/prod parity
• Decompose systems into cooperating services: Java backend + Python microservice + Helm-based Kubernetes deployment (see Briefly)
• Integrate security at the infrastructure layer — custom JWT filter chain, role-scoped endpoints, and admin registration gated by a shared secret header

Backend Java 17 · Spring Boot 3.x · Spring Security · Spring Data JPA · Python 3 · FastAPI · Node.js · Express.js

Databases PostgreSQL · MySQL · MongoDB · H2 (testing)

Infrastructure & Tools Docker · Docker Compose · Helm · Kubernetes · Maven · Gradle · Git · Caffeine Cache

Production-grade e-commerce REST API — cart, checkout, order lifecycle, payments, returns, audit trail.

Engineering Highlights

Impact Full e-commerce lifecycle in a single deployable unit: auth → catalog → cart → checkout (with promo codes) → order tracking → return requests → audit export. Documented via Swagger UI.

Multi-service URL summarization app: Spring Boot API + FastAPI LLM microservice + React UI, deployable to Kubernetes via Helm.

Engineering Highlights

• Service decomposition: Spring Boot (Java/Gradle) owns API routing and summary persistence; FastAPI (Python) isolates LLM inference — two runtimes, each independently scalable and replaceable
• Deployment: Helm chart for Kubernetes; root-level Docker Compose for local parity; each service ships its own Dockerfile
• Separation of concerns: The Java backend treats the LLM service as an internal dependency — callers never talk to the model server directly

Impact Demonstrates polyglot service design and cloud-native deployment readiness beyond a single Spring Boot monolith.

Automated aircraft damage detection system — built for Airbus Aerothon 6.0.

Engineering Highlights

• Fine-tuned YOLOv5 on a custom aviation dataset (Roboflow) to detect and localize cracks, dents, and deformities on fuselage and wings
• Random Forest classifier trained on image-extracted features to flag faulty wiring within aircraft harnesses
• Streamlit backend integrates model inference with file upload, result visualization, and repair recommendation output
• End-to-end pipeline: image upload → preprocessing (OpenCV) → damage classification → repair suggestion

Impact Working prototype delivered under hackathon time constraints; covers the full detection-to-recommendation loop for three damage categories.

Blood donation matching platform connecting donors, recipients, and blood banks.

Engineering Highlights

• JWT + Google OAuth2 via Passport.js — local and social auth strategies, session-aware cookie handling
• MongoDB/Mongoose schema design for donor profiles, donation requests, and bank inventory
• Express middleware chain: CORS → cookie-parser → auth → route handlers

Impact Full-stack MERN application with dual auth strategies and geo-aware donor matching flow.

I start with correctness before optimizing for performance. That means: define the contract, handle the failure modes, make operations safe to retry, then add caching or async paths where the profiling justifies it. I prefer explicit over implicit — if a service has a side effect (audit log, notification, stock decrement), that effect is visible in code, not hidden in an interceptor with no obvious trigger.

• Distributed systems fundamentals: consensus, replication, and partition tolerance trade-offs
• High-level and low-level system design (rate limiters, notification fan-out, distributed queues)
• Deepening Spring ecosystem knowledge: reactive streams, Spring Batch, Spring Cloud patterns