03 · How we learn
How we learn
Start with one real workload. Each build session improves a concrete layer of your stack, while the POD reviews architecture, trade-offs, and failure cases.
All courses
AI43 · Open models
Open Source & Self-Host
Build a private, capable AI stack around open LLMs: choose models, run them on infrastructure you control, and make each task earn its compute.
- Duration
- 4 weeks · 12 build sessions
- Format
- Telegram POD · self-host lab
- Pace
- Three build sessions per week
Waiting list
01 · What you’ll learn
What you’ll learn
- 01
Map privacy, latency, cost, and hardware constraints—and decide what belongs locally, in a private environment, or with an external provider.
- 02
Choose open models for dialogue, structured extraction, coding, retrieval, and multimodal work.
- 03
Deploy an inference gateway, route tasks between models, add useful context, and use caching to improve speed and cost.
- 04
Build evaluations for quality, safety, speed, and cost; observe failures and improve the stack without losing human control.
03 · Who it’s for
Who it’s for
- Teams and independent builders with a sensitive or recurring AI workload
- Product, operations, and IT practitioners who need control over data and tooling
- Advanced learners ready to run and evaluate models instead of only consuming hosted chat tools
06 · AI43 POD
Learner perspectives
“I stopped asking AI for answers and started building criteria for better decisions.”
“The POD made the difference: seeing how others approached the same quest exposed my blind spots.”
“Short quests fit around work, but the method stayed with me after the course.”
AI43 · Telegram POD
Run an AI stack you control.
Join the waiting list and we’ll share dates when the first Open Source & Self-Host POD is ready.
Get notified when the next POD opens.AI43 · POD 0418 members · AI coach online
AI43 · POD 0418 members · AI coach online