Utkarsh Pathak

Java • Built a complete interpreter for the Lox programming language implementing lexical analysis, recursive-descent parsing, AST construction, runtime evaluation, closures, lexical scoping, and control flow. • Developed a multi-stage language processing pipeline consisting of tokenization, parsing, AST generation, and interpretation, along with an interactive REPL for program execution.

TypeScript, Node.js, Redis, Kafka, PostgreSQL, Docker, React • Built a production-style API gateway featuring JWT authentication, API-key authorization, round-robin reverse proxying, and atomic Redis Lua-script token-bucket rate limiting. • Designed an event-driven analytics pipeline (Kafka → batch consumer → PostgreSQL), batching writes in 50-row/2-second flush windows to minimize database overhead. • Load-tested the gateway with k6 across multiple traffic scenarios (3,583 requests, up to 70 concurrent VUs), achieving 0% error rate and 23–41ms p95 latency. • Implemented fail-open Redis degradation handling and local Kafka buffering during broker outages to improve system resilience.

React, Node.js, PostgreSQL, Redis, Kafka, Docker, Kubernetes, k6 • Built a distributed URL shortener with Base-62 short-code generation supporting 3.5T+ unique URLs using collision-safe retry logic. • Designed a cache-aside architecture using Redis as an L1 cache in front of PostgreSQL, reducing redirect latency through in-memory lookups with graceful fallback on cache misses. • Implemented an asynchronous analytics pipeline using Apache Kafka, decoupling click-event logging from the redirect request path to keep redirects non-blocking under load. • Containerized services with Docker Compose and deployed to Kubernetes with Horizontal Pod Autoscaling (2–8 replicas) for traffic-based scaling. • Load-tested mixed redirect/create/analytics workloads using k6 (95 concurrent VUs), achieving 45ms p95 latency and 0% error rate across 7,500+ requests.