I design, evaluate, and productionize autonomous AI agents — from research and prototyping all the way to production and customers.

Currently a Senior AI Engineer @ DataCamp, optimizing one of edtech's top-rated AI tutors through evaluation science and A/B testing. Founder & CEO of Bareq, an agentic AI solutions agency. Founding R&D AI engineer @ Stakpak (Silicon Valley), building an open-source DevOps coding agent competing head-to-head with Claude Code and OpenCode.

• Agentic search, intelligent retrieval & context engineering
• Agent evaluation frameworks, benchmarking datasets & LLM-as-judge pipelines
• Autonomous coding & infrastructure agents (24/7 long-running systems)
• Multi-agent orchestration & dynamic memory systems
• Auto-prompt optimization (textual-gradient refinement)
• Hallucination detection & recovery pipelines
• RAG systems with trusted, fresh retrieval
• Agent observability, telemetry & production reliability
• Fine-tuning small LLMs (LoRA / QLoRA)

Optimizing DataCamp's AI-native course tutor, among the most efficient and top-rated AI tutors in edtech.

• Run sustained evaluation cycles — curating evaluation datasets that close blind spots, surfacing weak spots, and A/B testing across models and AI system designs to drive measurable quality gains.
• Shipped the platform guidance feature A→Z into production, full-stack from a Python/FastAPI backend to an Alpine.js + HTMX interface; added escalation detection and reporting for reactive AI behaviors.
• Built analytics & evaluation pipelines on GCP BigQuery to mine learner-interaction data, quantify tutor performance, and prioritize AI product improvements.

Founded and lead an agentic AI software agency delivering cybersecurity agents (SOC analysis), HR-department automation, R&D process automation, intelligent CRMs, and AI-native web apps — 5 clients closed, owning delivery from architecture to deployment.

Full-cycle AI product engineer on the founding team of an open-source autonomous CLI agent, implemented in Rust and Python.

• Architected an end-to-end agentic search & retrieval system (Rust) with caching, semantic reranking, trusted-domain filtering, and iterative LLM-based refinement — +40% recall, +55% precision with guaranteed freshness (benchmarked Crawl4AI, Playwright, Context7 MCP).
• Pioneered Stakpak's first DevOps evaluation dataset on the Harbor framework (Terminal-Bench standard) — containerized tasks with Docker environments and automated verification.
• Built an automated agent-testing framework with full A/B cycles: dataset generation, LLM-as-judge pipelines (Langfuse observability), and sub-goal trajectory analysis over 1000+ coding sessions.
• Designed an auto-prompt-optimization framework (inspired by DSPy & Adalflow) and hallucination detection/recovery via concurrent supervisor threads.
• Fine-tuning experiments on Qwen 3B (Unsloth, LoRA/QLoRA); co-developed Autopilot, the 24/7 autonomous infrastructure agent mode.

• Reliable autonomous coding agents
• Evaluation science for LLM systems
• Context & tool design engineering
• Prompt optimization algorithms
• Agent observability & telemetry
• Infrastructure-aware AI systems

SOC AI Analyst Co-Pilot — @ Bareq Self-hosted agentic co-pilot embedded in the SOC platform, wired into the SIEM, incident DB, and threat-intel stack; every verdict grounded in live tool calls against the client's own data. Delivers triage verdicts with evidenced rationale in under 30s (~90% faster time-to-triage) via parallel investigation. Security-grade: fully self-hosted models, role-aware intelligence tiers (Tier 0 → Tier 2 with MITRE ATT&CK mapping), strict memory isolation, redacted audit trail, human-in-the-loop by design.

Sarmad — AI Islamic Library (sarmadlib.com) — @ Bareq, Lead Engineer A "NotebookLM for hadith & Islamic literature": customizable notebook mode, full-database hybrid search, and voice-enabled reading (React, Django, Supabase, Python). Built OCR & scraping pipelines for digitizing books; hybrid retrieval (Typesense BM25 + semantic embeddings + Gemini long-context); multi-agent system with fast agentic RAG, dynamic prompt selection, summarization, and podcast-script generation.

Seqoon (hiresai.ai) — Founding AI Engineer Shipped SAI Dubai, a bilingual (AR/EN) real-estate chatbot on a full RAG pipeline (Milvus, Pinecone, Neo4j) with multi-turn LangChain/LangGraph agents; prototyped SAI Sales Agent, an Egyptian-Arabic voice AI (Twilio, LiveKit, Whisper/XTTS/Sesame, low-latency inference via vLLM & Ollama).

• AI Engineering Workshop Series — SOIL Coworking, Cairo · Lead Instructor & Organizer (Jun–Jul 2026, 4 sessions, 16+ teaching hours). A 4-session agentic AI series from GPT foundations to shipping production agents, with 8+ industry mentors, closing with a hackathon night of up to 25 pitching teams https://www.youtube.com/watch?v=7ieDMwmKSn4&t=151s .
• YouTube Live with Ahmed El Emam — Guest speaker on AI engineering, coding agents & agentic-AI careers.
• DevOpsDays Cairo 2025 — Technical speaker: live demo of the Stakpak agent dockerizing a web app in under 15 minutes.
• GenAI Hackathon Cairo 2025 — Co-organizer & workshop leader (150+ participants); recruited judges from Vectara, Valeo, and Thundr.

Python • Rust • Modern C++/C • Java • SQL • Bash • VHDL

LangChain • LangGraph • Langfuse • DSPy • Adalflow OpenAI • Ollama • vLLM • Unsloth • LoRA/QLoRA fine-tuning Context engineering • MCP • Multi-agent systems Agent evaluation (Terminal-Bench / Harbor, LLM-as-judge)

TensorFlow • Keras • PyTorch • Scikit-Learn • Pandas • NumPy • NLTK • OpenAI Gym

Milvus • Pinecone • Neo4j • Typesense Hybrid search (BM25 + embeddings) • Semantic reranking • Metadata-aware RAG

FastAPI • Axum • Django • Alpine.js • HTMX • React • Streamlit Supabase • PostgreSQL • MongoDB • GCP BigQuery

Docker • Kubernetes • K3s • Harbor • Traefik • HAProxy • NGINX Cloudflare Workers/Tunnels • AWS • GCP • Coolify • CI/CD

Prometheus • Uptime Kuma • Trivy • Checkov • k6 • perf • memray • SIEM integrations (SOC agents)

Git • GitHub • Linux • Twilio • LiveKit

Spoken: English (Full Professional) • German (A2) • Arabic (Native)

• LinkedIn: https://linkedin.com/in/abdalla1912mohamed
• Stakpak Agent: https://github.com/stakpak/agent
• Bareq: https://bareq.dev
• Sarmad: https://sarmadlib.com

Beyond the terminal: Military parachuting & skydiving — static-line jumps from C-130 aircraft and Mi-17 helicopters as Squad Leader. 🪂