const developer = {
name: "Roshan Raj",
location: "IIT Bhilai, India",
education: "B.Tech in Computer Science & Engineering (Expected 2027)",
focusAreas: ["Systems & Compilers", "Distributed Systems", "Full-Stack Dev", "Machine Learning"],
techArsenal: {
languages: ["C++", "Go", "Python", "TypeScript", "JavaScript", "SQL"],
frontend: ["React", "Next.js", "Tailwind CSS"],
backend: ["Node.js", "Express.js", "Socket.IO", "PostgreSQL", "MongoDB"],
machineLearning: ["PyTorch", "scikit-learn", "XGBoost", "LightGBM", "OpenCV"],
systems: ["Linux", "Docker", "WebAssembly", "GitHub Actions"]
},
research: "S3 Summer Labs intern: ML + embedded software for NISHAD, an anemia-screening device (with AIIMS Raipur)",
openSource: "5 merged security & privacy PRs to OpenLake's RateMyCourse",
competitiveProgramming: "350+ problems solved on LeetCode",
philosophy: "Building the future, one line of code at a time."
};
βοΈ MiniLangA compiler and stack-based bytecode VM written from scratch in C++17: lexer, recursive-descent parser, semantic analyzer, optimizer, and a 60-opcode VM, compiled to WebAssembly for a browser IDE with a step-through debugger. |
A Go reverse proxy with least-connections routing, active health checks, and retries across 3 replicated backends. Traced an 18% error rate to OOM kills and brought it to 0 errors at 2,500 concurrent users. |
β‘ AlgoRace1v1 real-time coding races: ELO-based matchmaking, live opponent progress over Socket.IO, 102 problems with 1,020 test cases, and single-winner settlement in one database transaction. |
An active learning workflow using Bayesian Optimization and Gaussian Process models to speed up the discovery of high-entropy thermoelectric materials. |

