Harshit Thakkar

GPA: 3.7 / 4.00

Coursework: Distributed Systems, Advanced Data Structures, Algorithms, Software Architecture, Parallel Computing

GPA: 8.2 / 10.00

Coursework: Object-Oriented Programming, Operating Systems, Database Management, Machine Learning

Blueinfy · Ahmedabad

• Engineered a custom JavaScript Concrete Syntax Tree parser achieving 98% accuracy, integrated with Semgrep's workflow to automate static code analysis and review code quality across 100k+ lines of Gosu.
• Applied rigorous live debugging to diagnose complex technical bottlenecks in *nix environments, delivering actionable, well-documented logic that enhanced security governance and reduced manual audit cycles by 25%.

Level Supermind · Mumbai

• Consolidated freeze streak and regular streak microservices into a single database structure by modeling the freeze streak as a standard activity, streamlining PostgreSQL pipelines to ensure continuous data availability.
• Analyzed and optimized core algorithms to reduce computation from quadratic to log-linear time, slashing latency from 1.2s to 150ms to strictly fulfill a 1-second API contract for iOS frontends.
• Packaged scalable architectures via Docker to reliably support 10,000+ concurrent requests, establishing end-to-end ownership of observability metrics to drive a 30% increase in user engagement.
• Maintained comprehensive operational runbooks for live debugging and 24/7 support, ensuring seamless adoption across global engineering squads.

• Collaborative infinite-canvas mind-map with a WebGL frontend, quadtree-backed spatial engine, and local-first map storage on the user's device.
• Spring Boot REST API with Google OAuth and MongoDB; documented GCP and DigitalOcean deployment workflows.

• Desktop map database suite with a custom .plusmap binary format, quad-based spatial indexing, and import/query engine for terrain and road features.
• Core and engine modules complete with 47 passing tests; powers the spatial backend behind Brain Map.

• Local-first YouTube content filter using offline LLMs to classify recommendations against a user profile and block irrelevant tiles — zero cloud dependency.
• JavaFX desktop app controls the Chrome extension via localhost IPC, with profile sync, live stats, and sub-180ms classification latency.

• Multiprocess code execution platform running Java, Python, C++, JavaScript, and Bash in isolated OS processes with real-time stdout/stderr streaming.
• Enforces per-run time and memory limits with TLE killing, live process history, and Chaos Monkey compatibility for resilience testing.

• Built a Simian Army suite for a university distributed-systems course — chaos, latency, conformity, and load monkeys to stress-test a social-media microservices app under simulated production failures.
• Instrumented services with observability tooling and validated resilience patterns; monkey scripts live in the public repo — the full university social-media app stays private.

• Swing GUI that solves scrambled cubes using the beginner's method, with text-scramble and per-face color input modes.
• Animates the solve step-by-step with pause, rewind, speed control, and move notation highlighting for learning standard cube notation.