Currently · SDE II at Niyyah
Omar Salama
React Native & Python. RAG pipelines, on-device LLMs, five years of shipped products.
5 yrs shipping mobile · Production RAG & agentic systems · On-device LLM fine-tuning
Open to remote (CET/EU) & relocation · Giza, Egypt
About
Software engineer, five years in. Two tracks that keep overlapping: AI systems and mobile.
On the AI side I build production LLM systems end to end: retrieval and RAG (hybrid dense and sparse search, rank fusion, cross-encoder reranking, multilingual indexing), agentic orchestration that splits routing from generation (a router gathers and ranks sources, a separate model grounds the answer with citations), model fine-tuning (SFT and DPO with LoRA), and the evaluation tooling that keeps all of it honest.
On the mobile side I started in native Android and iOS, moved to React Native, and shipped cross-platform features for years, including deliberate UI/UX redesign passes that brought apps from a working state to a considered one. Most recently, a fine-tuned LLM running fully on-device in a React Native app.
Three and a half years at Keepoala, just over a year as a full-time LLM-training contractor at Turing. Now an SDE II at Niyyah, working across the AI backend and the React Native app.
A few other things along the way: NeuroGames, a VR brain-controlled game (Cairo University graduation project). An object-aware image compression algorithm. Optical music recognition, from sheet music to MIDI. A React and R-Shiny analytics dashboard for partner shops.
On the side I built OpenSpace, an on-device therapy companion that runs a Gemma-4 2B model I fine-tuned and published as MindSpace on HuggingFace, with on-device retrieval and a two-stage SFT plus DPO pipeline behind it.
“The work I like best runs the whole length of a feature: the model, the retrieval that feeds it, and the app it ships in.”
Experience
Sep 2025 - Present
Niyyah
SDE II
Work spans the AI backend and the React Native app. I build the retrieval-grounded answer pipeline (a router gathers and ranks sources, then a separate model generates a cited, streamed answer, with model routing and fallback for quality, cost, and reliability) and, from scratch, the evaluation tooling that scores it (LLM-as-judge scoring, retrieval-quality metrics, regression checks with alerting). I cut input tokens per answer by 10–20%, and made responses resilient with a server-side streaming proxy that persists in-flight answers, so they survive the app closing or losing connection, plus conversation summarization that keeps long chats within budget. I also plan features end to end, like a personalized 30-day guided journey or a conversational onboarding flow, then ship the React Native side.
Jun 2024 - Sep 2025
Turing
Python Developer (LLM Training)
LLM training contractor; SFT for domain-specific applications. Promoted to pod lead, guiding a 4-person pod and reporting on performance; built internal Jupyter tooling for cross-pod progress reports.
May 2021 - Oct 2024
Keepoala
Software Developer
Core mobile engineer on a two-person mobile team; built and shipped the React Native consumer app (iOS + Android) over 3.5 years, from working student onward. Partner-shop analytics dashboard (React + R-Shiny), Firebase integrations, BillBee/WeClapp/Shopify connectors.
Oct 2021 - Mar 2022; Jan 2025 - Apr 2025
Karat
Interview Engineer
Two contract stints. Delivered 10-15 structured technical interviews for software engineering candidates on behalf of Karat's enterprise clients, using calibrated rubrics.
Projects
Skills
Mobile
LLMs / Fine-tuning
RAG / Retrieval
Agents
Evaluation & observability
Backend & Infra
AI-assisted dev
Languages
Contact
Let’s talk.
Looking for mobile or AI engineering roles, remote or in Egypt. Email me and I’ll usually reply the same day.