I'm Sanish Kumar—a Full-Stack & AI Engineer blending Python, Go, and React expertise. I build scalable microservices, generative AI pipelines, and interactive front-ends with brutal efficiency.
I'm a Computer Science student at Lovely Professional University (Class of 2026), obsessed with structural integrity in code. I reject bloated, inefficient systems in favor of high-performance microservices and seamless front-end experiences. My stack is unapologetic: Python, Go, React, and Next.js.
When I'm not coding, I'm competing in hackathons, exploring new AI research, and building generative AI pipelines. Every line of code must justify its existence.
A distributed fintech platform that optimizes complex group debt networks in real-time. It features a high-performance C++ WebAssembly (WASM) Max Heap solver, a React Flow interactive visualization, and a scalable backend powered by Node.js, PostgreSQL, Socket.io, and Redis Pub/Sub.
Production-grade AI system bridging deterministic AEC engineering data with multi-agent orchestration to analyze raw .ifc (Industry Foundation Classes) files. Engineered a robust Python pipeline utilizing IfcOpenShell to programmatically extract exact 3D structural geometry, bypassing LLM spatial hallucinations. Deployed specialized AI agents via CrewAI to evaluate extracted base quantities against building codes and pricing matrices, delivering highly accurate real-world engineering documentation and cost analyses.
Built an interactive sketch-to-image web app that lets users draw rough doodles on a canvas and transform them into polished illustrations using an AI model. Designed the UX and generation flow so that even messy or low-skill sketches can be converted into clean drawings through a simple draw-and-generate interaction, deployed on Vercel for fast, reliable access.
A JavaScript library that adds powerful voice-control capabilities to web-based maps. Uses a hybrid architecture routing between fast online cloud STT and robust on-device AI (Whisper/TF.js) for seamless offline fallback.
Engineered a real-time collaborative cloud architecture whiteboard utilizing React Flow and Socket.io, backed by a scalable Node.js microservice and Redis pub/sub for low-latency state synchronization. Integrated a pluggable LLM adapter to analyze complex architectural graph states (JSON), returning actionable engineering feedback and auto-generating missing infrastructure nodes directly onto the canvas.
Chronological
Data Sequence
INIT: 2026.05
Collaborated on R&D for NiftyIP, co-developing Python-based analytical pipelines to detect intellectual property (IP) infringement and unauthorized style replication across generative AI models. Contributed to core research in dataset auditing and membership inference, analyzing latent space representations. Worked within a cross-functional team to translate complex AI research into production-grade systems using FastAPI and advanced vision workflows (ComfyUI/Flux).
Orchestrated the integration of bounded and unbounded AI pipelines into a flagship web platform, developing highly robust client adapters for endpoint ingestion. Collaborated to break down complex system design requirements, delivering a 25% enhancement in accessibility and automated data routing.
Built a robust end-to-end backend system for a restaurant company using Python, MongoDB, and TypeScript. The work improved load speeds by 40% and cut senior team workload by 60% through Figma-driven UI mockups and clearer handoffs.