Applied AI Architecture Training
5 weeks · 20 hours · Free · Open to all · 20 Principal + 60 Reserve
2 evenings/week · 2 hours per session · applications until end of June · starts early-to-mid July
About the program
This is an intensive 5-week production camp that teaches how to build a holistic AI architecture starting from first principles. The goal is not just theory; from week one you contribute lasting work to the HuggingFace ecosystem by shipping datasets, models and live apps (Spaces). Each week turns theory into a concrete output, leaving you with a real portfolio.
Schedule & participation
- Format: 2 evenings per week, 2 hours each (20 hours total over 5 weeks). Evening slots are chosen so most people can attend after work or school.
- Applications: Open until the end of June.
- Start: Early-to-mid July. The exact date is shared with accepted participants.
- Setting: Fully online and live; all sessions are recorded.
Process & fair evaluation
- Automated pre-review (Review Bot): When you push assignments to GitHub or upload to HuggingFace, repo format, code structure and required outputs are checked automatically — so feedback is fast and consistent.
- Transparent leaderboard: A public ranking updated by assignment speed and quality. Promotions to the principal list follow this ranking directly.
- Constructive participation metric: During live sessions, AI moderation evaluates not talk time but NLP metrics such as terminology use, topical relevance and constructive contribution. The goal is real contribution, not empty questions.
Prerequisites
- →Aktif HuggingFace hesabı
- →Aktif GitHub hesabı
- →Google Colab arayüzüne aşinalık
- →Temel Python bilgisi
Curriculum
Anatomy of tokenizer architecture and core LLM principles
Open-source data ecosystem and domain-specific dataset culture
Create a domain-specific dataset and publish it on HuggingFace in a standard format.
Rules & System
Rules & Information
Application prerequisites: Active HuggingFace + GitHub accounts and Colab familiarity. Profile links are required in the application form.
Principal/Reserve dynamic capacity: Principal (20) join live sessions; Reserve (60) get recordings right after class and must complete the same assignments on the same schedule. Performance drives movement between lists.
Assignments & attendance (zero tolerance): One missed assignment moves a principal to the bottom of the reserve list; the top-performing reserve is promoted. Three missed assignments remove a reserve from the program.
Live session rules & AI moderation: Camera required; passive or disruptive participants may be removed from the principal list via AI moderation. Moderation is based not on talk time but on NLP metrics such as terminology use, relevance and constructive contribution.
Certification: Everyone who completes the program in compliance with the rules receives a verifiable certificate.
Outcomes
- ✓HuggingFace ekosisteminde veri seti, model ve Space yayınlayabilme
- ✓Fine-tuning (LoRA/QLoRA) uygulayabilme
- ✓Tool call ve MCP entegrasyonu yapabilme
- ✓Embedding modeli seçimi ve vektör veritabanı kurulumu
- ✓Uçtan uca RAG sistemi tasarlayabilme
- ✓Doğrulanabilir Başarı Sertifikası
Everyone who completes the program in compliance with the rules receives a verifiable certificate.
Apply to this program
Verify your Google account and complete the form.