I ship AI to production. Most consultants only demo it.
AI Systems Engineer. Founder of NeuraScale. 9+ years engineering. Egypt (GMT+2), covering EU mornings and US afternoons.
I build AI systems for real businesses and run them in production after launch: retail, medical exam prep, B2B sourcing, salons, cafes. Five products live right now. I own the whole path: data model, backend, UI, deploy, monitoring.
Two tools I maintain and use in my own production work:
- llm-eval-ci: a CI quality gate for LLM products. Golden-set regression tests with calibrated graders (grounding, hallucination, tool calls, LLM judge); the PR fails when answer quality drops. About 600 lines, one dependency, MIT.
- mnemonic: self-hosted layered memory for AI agents. Tiered context tree (L0/L1/L2 summaries), auto-capture and recall on every turn, contradiction resolution. Runs entirely on your own server, MIT.
- 5 live products: MedPrüf, RetailOS, Bridge Sourcing, Harmonia POS, Crema.
- 10,992+ exam questions live on MedPrüf, exam prep for foreign-trained doctors in Austria.
- Empty repo to live cafe POS in 1 day: Crema went live on its production URL the same day it was started.
Entry point: Find the leak. $950, one week. You tell me what feels slow, manual, or expensive in your business. You get a plain-English plan (what to fix, in what order, what it costs) plus one real fix, built and working. Book it at omargnagy.com/audit.
Full builds and ongoing partnerships: omargnagy.com.