🚀 Agentic AI Workshop Series — Session 1: Build Your Own GPT
🚀 Agentic AI Workshop Series — Session 1: Build Your Own GPT
A 4-session intensive on advanced agentic AI practices. Fully self-funded, fully free, for a hand-picked cohort of highly talented senior students and engineers.
By the end of these four Saturdays, you leave as a practitioner — not a student.
📅 When & Where
📍 Soil Co-working Space, Mivida — Orange Meeting Room
📅 Starts Saturday, 20 june 2026
⏰ Every Saturday (or biweekly), 6:00 – 10:00 PM (4 hours intense)
Potential session dates ( adjustable) :
Session dates:
• Session 1 — Saturday, 20 June 2026
• Session 2 — Saturday, 4 July 2026 (2-week gap)
• Session 3 — Saturday, 11 July 2026
• Session 4 — Saturday, 18 July 2026 — Hackathon night
💰 Totally free — sponsored by me
🎯 Who This Is For
This is not a beginner workshop. We move fast, go deep, and expect you to keep up. Built for senior students and working engineers who want to actually understand what’s under the hood of modern agentic systems and ship real ones.
Reserved seats for partner startups: Akedly, Raid.AI, perceptr.io.
✅ Prerequisites
Before you apply, make sure you have:
• Solid Python basics — comfortable with classes, generators, type hints, virtual envs
• Some AI knowledge — you’ve called an LLM API, you know what a prompt is, you’ve seen pip install transformers
• ML background — a plus, not required. Familiarity with neural nets and gradient descent will help in Session 1.
• Working laptop + charger — Mac, Linux, or Windows all fine
• Real intent to learn and build — this is the only non-negotiable one
If you’re missing any of the technical pieces, this isn’t the cohort for you yet — but follow for the next one.
📚 The Series
Session 1 — The Foundations: From Tokens to Agents — Build Your Own GPT
NLP, embeddings, attention, decoder-only transformers. We build a small GPT from scratch in TensorFlow, train it on text + code, and close on modern LLM architecture and RAG.
Session 2 — The Agent Stack: Tools, Context, and Control — Build Your Own Coding Agent
Tool calling, MCP, context engineering, prompt caching, hallucinations, and guardrails. We build a coding agent with agentic search that explores a codebase and ships a fix.
Session 3 — Production Agents: Evals, Memory, and Self-Improvement — Build Your Own CLI Agent
End-to-end CLI agent with evals, hallucination detection, automatic prompt optimization, memory, and real system design — while reading the source of OpenCode, Stakpak, and Open Claude Code.
Session 4 — The Agent Hackathon — Build, Defend, Discuss
Teams ship working agents. Mentors tear them apart constructively. We close with an open discussion on where agentic AI is actually heading.
👥 The Lineup
Led by:
Abdalla — ex-founding engineer R&D @ Stakpak (Silicon Valley), ex-founding AI engineer @ Seqoon, ex-TA at The German University in Cairo.
Mentors & Practitioners:
Mostafa ashraf — ex-Stakpak , ex-Reflexer Labs , CTO & Co-Founder of 3BONT , 15+ years in software development
• Ahmed Sallam — Tekunda, ex-Stakpak, ex-Orange. 15+ years in product and growth.
• Abdelrahman Soliman — Founder @ perceptr, Senior Software Engineer @ Tether.io.
• Muhab Samir — CEO @ Akedly.
Organizers:
Abdalla · Ahmed Tamer (Senior Engineer @ Etisalat) · Mohamed Samy (Software Engineer @ Procore Technologies).
A lineup spanning Silicon Valley, deeptech startups, and senior engineering across Vois, Etisalat, Procore, and more.
📜 Roles & Expectations
We run this seriously because we’re investing serious time. Here’s what each role looks like:
🎓 Participant
• Full attendance — miss one Saturday, you’re out for the rest
• On time — 20-min grace period, once only
• 4–8 hours of homework + reading between sessions
• Active in-session participation: laptop open, building, asking
• Stay all 4 hours, every session
• Respect mentors and peers
🤝 Community Partner / Sponsor
• Help cover AI API credits and/or co-working space costs
• In return: branding across announcements, posts, and session materials
• Access to a curated cohort of serious AI builders
• Recognition as a partner of the AI community in Egypt
🎤 Hackathon Judge (Session 4)
• Evaluate team agent demos
• Critique architecture, evals, and design choices
• Join the closing open discussion on the future of agentic AI
📰 Media Partner
• Coverage of the series, mentors, and participant stories
• Access to interviews and live workshop coverage
🎒 What to Bring
• Laptop + charger
• Notebook + pen
• Open mind
• 4 hours of full attention
📝 How to Apply
RSVP on this Luma page — and fill out the application form linked in the comments. Seats are interview-screened. We’ll review and confirm within 48 hours.
🤲 Open Asks
Looking for community partners to help fund AI credits and venue for the next cohort, and media coverage for the series. Hackathon judges for Session 4 also welcome. DM Abdalla if interested.
🚫 Rules — Non-Negotiable
1. Full attendance. Miss one = removed from the series.
2. Be on time. 20-min grace, once. Late again = out, permanently.
3. Homework is mandatory. 4–8 hours between sessions. Not done = out.
4. Active participation. Passive attendees removed mid-session.
5. Respect your mentors. They’re giving their time for free.
6. Stay the full 4 hours. Leaving early = same as missing.
We picked you for a reason. Show up, build, and carry it forward.