Shared AI services
ShortsBlink: one model gateway reused across applications.
AI prototype stuck in review?
Your next AI engineer
already works here.
Train the developers you already have to build AI features they can test, ship and maintain.
Learn from production experience
I’m Ion Moșnoi.
ML engineer. Product builder. Your instructor.
My experience on LinkedIn ↗8+ years in ML and software. Dozens of projects in production.
I bring the decisions behind those systems into your team’s work: what failed, how we measured it and what we changed.
Evaluation & correction
LLM Arena and TennisLens: compare outputs, review mistakes, improve the next version.
What your engineers learn
Find the failure. Know what to change.
Get the data right
Clean, label and version data. Understand splits, leakage and when model training helps.
Fix retrieval
Choose RAG and chunking strategies by testing which evidence reaches the model.
Test the task
Build eval datasets and metrics that expose wrong answers and regressions.
Control agents
Build a harness with tools, state, retries, limits and human approval.
Improve with evidence
Compare data, prompts and models. Turn reviewed mistakes into better tests.
Ship maintainable code
Use AI coding tools with review and tests. Add logs, monitoring and reusable interfaces.
How the workshop works
Your feature. Your stack. 1–2 days together.
Choose one problem.
We review your code, approved examples and release blockers.
Build, test, improve.
Your engineers run the work, inspect failures and compare a change.
Keep building.
Leave with reference code, starter evals, a checklist and a clear next sprint.
We choose the relevant modules for your team. One day tackles a focused issue; two days allow a deeper exercise.
Bring the AI feature that’s stuck.
Tell me what your team is building and where it breaks.
Talk about your project ↗