Hi, I am Fatin, a software engineer building AI agents.
CS at UMBC, graduating December 2026. Lately I've been writing an LLM inference engine in C++ and measuring exactly why it's slower than llama.cpp.
0.0 tok/sLLM Engine on an M4
0% fastercontract review, in testing
0+ runsvalidated automatically
0 agentscoordinated in LectureKit
Experience
Built a Python tool that checked 60+ quantum simulation runs against reference solutions, cutting manual review from 4 hours to 5 minutes.
Education
B.S. Computer ScienceUniversity of Maryland, Baltimore CountyExpected Dec 2026Expected Dec 2026
Projects
- Python
- C++
- TypeScript
- Next.js
- FastAPI
- PostgreSQL
- Docker
- LangChain
- MCP
Latest writingWhy llama.cpp generates text almost 9x faster than my engine ↗Benchmarks on an Apple M4, where the time goes, and what I'd fix first.
Contact
Hiring for a new grad role, or want to talk about a project? Send a note here or email fatinm1@umbc.edu.