Catalog · UpSkillZone AI
Three tracks. Each ships a portfolio a hiring manager can read, run, and verify.
UpSkillZone AI runs a small number of vertically-deep tracks instead of an endless catalog. Each is a 10–14 week cohort with a real capstone artifact, a calibrated mentor attached, and a public credential at the end. The point is not exposure to a topic — the point is to ship work that survives review.
Every credential below is W3C VC + Open Badges 3.0, signed by the issuer key, indexed in the public skills directory. If you do not place against the stated outcome inside the published window, the outcome credit is refunded — that clause is in the enrolment agreement, not a footnote.
Open for enrolment
1 live
production-ai-engineer · Cohort 04 opens 2026-06-03
Production AI Engineer
A 14-week program that turns a working software engineer into a Production AI Engineer.
- Duration
- 14 weeks · 12–15 hrs/wk
- Prereqs
- 2+ yrs shipping production code, comfort with Python and HTTP
- Credential
- Production AI Engineer — W3C VC + Open Badges 3.0
- Next start
- Wed, 3 Jun 2026
On the roadmap
2 planned
Tracks below are spec'd and currently in build. We publish their cohort window before enrolment opens so the roadmap is legible, not aspirational.
llm-eval-engineer · Cohort 02 opens 2026-Q3
Syllabus locked, mentors signed
LLM Eval Engineer
Capstone: a calibrated judge harness — golden sets, inter-rater agreement floors, and CI-graded regression dashboards for a production LLM feature.
- Duration
- 10 wk · 10–12 hrs/wk
- Prereqs
- Shipped at least one LLM feature; comfort with pytest + GitHub Actions
- Credential
- LLM Eval Engineer — W3C VC
production-ml-ops · Cohort 03 opens 2027-Q1
Spec in review
Production ML Ops
Capstone: a multi-region training pipeline — feature stores, reproducible runs, drift detection, zero-downtime model rollouts.
- Duration
- 12 wk · 12–14 hrs/wk
- Prereqs
- Working knowledge of Docker, one cloud provider, and a model lifecycle tool
- Credential
- Production ML Ops — W3C VC
What we won't ship
- No 4-week "become an AI engineer" track. Production competence does not compress to a month.
- No "intro to" tracks. The catalog is for people who already write software and want to ship a specific thing.
- No certificates without a signed credential. If a hiring manager cannot verify it from the URL, we did not issue it.
Background reading
The thesis behind picking three tracks instead of three thousand is on the manifesto.
Verify what we issue
Every credential these tracks issue is queryable in the public ledger.