I’m a computer engineering student and full-stack developer who enjoys building things across the stack, from front-ends and TypeScript/Node.js services to Python-based data and ML projects. I’m also interested in what happens under the hood, especially Linux, networking, and digging into code to understand how things actually work.

LinkedIn · linkedin.com/in/umutcanakcay  |  E-mail· can.akcay@aol.com

• Languages: JavaScript/TypeScript, Python, C, C++, Java, Bash, SQL
• Backend: Node.js/Express, NestJS, Flask, .NET MVC Core, REST APIs, WebSockets (Socket.IO)
• Frontend: React, Next.js, Vue, Tailwind CSS, Bootstrap
• Data & ML: pandas, NumPy, scikit-learn, LightGBM/XGBoost, embeddings and vector search (RAG), LangChain, LLM tool-calling
• Systems: Linux, Docker, WebRTC, P2P/DHT, parallel computing
• Spoken: Turkish (native), English (B2), German (A1)

• peercomp — serverless JavaScript on a Kademlia DHT over WebRTC: browser tabs become compute nodes, sandboxed and signed, with no cloud dependency.
• sessionflow — self-hosted, privacy-friendly web analytics: a tracking snippet feeds an Express/Socket.IO + MongoDB backend, a React dashboard shows the traffic, and a Flask service runs scikit-learn analysis.
• threat-aggregation-intelligence-platform — threat-intelligence IOC aggregation: normalizes ThreatFox and PhishTank feeds, deduplicates them, and serves search, MCP tools and AI-written reports.
• local-rag-ingest — local-first toolkit for a D&D 5e game master: campaign notes into a Qdrant vector store with Ollama embeddings, game state in SQLite, action data prepared for a local llama.cpp model. The tool-calling engine that consumes them is in progress.
• web-recon-scanner — passive web attack-surface scanner with a Streamlit UI: DNS, TLS and header checks, certificate-transparency subdomains, gau/theHarvester/whatweb.
• image-optimization-service — an image-optimisation API (Node.js/Express, MongoDB, Swagger) that compresses uploaded profile images above a size threshold and serves a small token-authenticated user directory.
• lean-mcp — an MCP server that compiles Lean 4 locally with the real toolchain and returns structured compiler diagnostics, so coding assistants can verify proofs (GPL-3.0).
• consultant-specs — document-first planning framework for AI coding agents: PRD, ADR and SPEC documents plus seven structured analysis tools. npx skills add HyoAkira/consultant-specs
• logi-m575-sensitivity-app — a small C/Linux utility that maps the extra buttons of a Logitech M575 trackball to pointer sensitivity through xbindkeys (GPL-3.0).