The Product Engineer
Hi, I'm Prajyot. I build software that works, not demos that impress. I have engineered AI tools like DevFlow and ContextGraph, also solved 417 LeetCode problems, and merged a production fix in JupyterLab, and currently freelance for global clients.
Read the PrinciplesProof of Work.
DevFlow
ContextGraph
Job Tracker
DevFlow
AI code assistants produced generic code due to slow, unoptimized context parsing.
Engineered a Gemini self-repair validation layer and optimized WebContainers WASM structure.
ContextGraph
Multi-tenant context engines face concurrency and memory bloat under scale.
Built a JSON-RPC 2.0 MCP server with PostgreSQL tenant isolation and 8% decay cron.
Job Tracker
Applying to dozens of roles without a central system leaves application trends and response patterns completely invisible.
Built a background email-scraping AI agent cron to parse confirmations and push structured data to Supabase.
Professional track record and academic baseline.
Engineering beliefs developed in active production.
Write code that can be deleted easily.
Software requirements evolve rapidly. By building highly modular, self-contained files and avoiding early dry-run abstractions, we ensure that obsolete features can be completely purged from the codebase without leaving zombie code or creating hidden dependencies.
Uptime is a product feature, not an operations metric.
Reliability directly impacts user trust and conversion rates. Implementing fault-tolerant multi-model LLM chains, self-healing code compilation, and zero-trust API layers prevents failures from ever reaching the client interface.
Static is always faster than dynamic.
Optimize for compile-time rendering. By utilizing Next.js Incremental Static Regeneration (ISR) and aggressively caching database queries, we deliver sub-second response times and bulletproof hosting scalability.
Abstract only when you have duplicated three times.
Early abstraction is far more expensive than duplication. Writing repetitive code in the short term keeps the architecture flexible, allowing the true structural pattern to emerge naturally before refactoring it into a library utility.
UX polish is a performance optimizer.
Intelligent micro-interactions, magnetic elements, and spring-driven animations decrease perceived latency. Instantly reacting to user inputs within 16ms keeps the product feeling snappy, even when waiting on slow network payloads.
Quantifying development speed and algorithmic depth.
Open Source Activity
Contributions to core engines, packages, and developer workflows. Plotted live from GitHub.
Algorithmic Baseline
Solved 424 problems (32 Hard) verifying computational complexity. Plotted live from LeetCode.
Project Jupyter / JupyterLab
Patched font size handler editor bug, preventing client-side application freezes. Merged PR #18157.
Merged PR #18157↗Competitive Engineering
- Finalist at ADCET 2024 and DYPDPU 1.0 2025.
- Qualified for the Avishkar State Championship.
Let's build something permanent.
I am open to engineering roles, consulting engagements, and technical collaborations. Let's talk about systems, products, and codebase architecture.