Git for Team Workflow
Branches, rebase vs merge, PR hygiene, recovery. The workflow every later week depends on.
Junior Track · Cohort 1 · Founding · Live Zoom
The AI Infrastructure Junior Engineer Academy.
16 weekly live sessions. 24 hours of instruction. One real production-shaped ML service shipped to your GitHub by Demo Day. The first program in a planned ladder of role-based AI infrastructure courses (Senior, MLOps, Architect, more — based on what the cohort says next).
Cohort 1 · Founding
$199
limited seats · price doubles for cohort 2
Cohort 2+ · Standard
$399
~$25 per week of live instruction
Why this Academy
Built for the parts of the job nobody else teaches.
Most ML courses teach you to train a model. Most infra courses ignore that ML breaks every assumption infrastructure makes. This program lives in the gap — the production shape of an ML service from week one.
01
One project, sixteen layers
You'll build the same sentiment-classification service every week, adding one production layer at a time — containerization, Kubernetes, CI/CD, monitoring, drift detection, pipelines, IaC. By week 16, you can point at it in interviews.
02
Code-along, not slides
Each session has 25 minutes of live coding where you build alongside the instructor. Not "watch me demo." Build it on your machine while we move together. The Academy feel comes from leaving each session having shipped.
03
Anti-fluff curriculum
Topics make the cut if they appear in >60% of real Junior AI Infra job postings. What got cut: theoretical ML, deep CS fundamentals, fashionable-but-rare tools. What stayed: Git workflow, Docker, K8s, FastAPI serving, Prometheus, Postgres, AWS + IaC, observability, ML pipelines, LLM serving.
04
Personal capstone review
Between weeks 14 and 15, you'll get a 30-minute personal code review on your capstone PR. Real line-level feedback from the instructor — the kind that makes your code interview-ready. Most cohorts at this price don't do this.
What's included
$25 a week buys you more than the live session.
Everything below is part of the program. No upsells, no add-on tiers — you either have access to the cohort or you don't. The $199 founding price covers all of this.
▶
16 live Zoom sessions
Weekly, 90 minutes each. 24 hours of synchronous instruction.
↻
Lifetime recordings
Miss a session, rewatch any time. Indexed with chapter markers.
?
Weekly office hours
One hour drop-in Zoom every week with the instructor. No agenda — bring blockers.
✓
Capstone PR review
30 min of personal line-by-line code review on your capstone before Demo Day.
★
Demo Day
Present your project to the cohort in week 16. Recorded, shareable, your graduation moment.
#
Private Discord
Cohort-only channels. Weekly homework threads. Producer + instructor present.
∞
Alumni community
Stays open after graduation. Cohort 1 alumni get the founding-member badge.
⬡
Cohort certificate
Signed PDF + a LinkedIn badge. Founding Member of Cohort 1, if you want it.
The 16-session arc
Each week, one production layer goes on the service.
Sessions are sequenced so each layer builds on the one before. By week 8 you have something running in K8s with CI/CD and monitoring — that's the "I shipped a real thing" moment. Weeks 9-15 add depth; week 16 graduates.
Production Python
uv, FastAPI skeleton, pytest. Code shape that survives review, not notebook patterns.
Linux & Shell
Filesystem, processes, pipes, journalctl. The 20 commands that buy you 80% of the leverage.
REST APIs & Model Serving
The real model goes in. Loading, versioning, contract design, failure modes.
Docker for ML
Multi-stage builds, layer-cache discipline, GHCR. ML images shouldn't be 4 GB.
Kubernetes, Just Enough
Pods, Deployments, Services, HPA. The mental model — not the certification.
CI/CD for ML
GitHub Actions. PR runs tests, main builds and pushes. OIDC over PATs.
Monitoring: Prometheus + Grafana
RED metrics, PromQL fundamentals, alerting that doesn't page on noise.
ML-Specific Monitoring
Prediction logging, drift signals, ML-specific failure modes — the parts traditional APM misses.
Logs & Observability
Structured logs, correlation IDs, OTel tracing. One ID threaded end-to-end.
SQL & Databases
Postgres, indexes, EXPLAIN, transactions, connection pooling.
ML Pipelines
Airflow DAG that retrains weekly. MLflow tracking. Idempotency patterns.
AWS + IaC
Terraform for ECR + EKS + RDS. IAM the way it's actually written. Cost levers.
End-to-End Capstone
Integration: trace one request through all 7 layers. Capstone PR opens.
Role + Stack
What the Junior role actually looks like. Capstone polish for Demo Day next week.
LLM Serving + Demo Day
45 min vLLM / batching / cost economics. 45 min cohort presents their capstones. Graduation.
Pricing
$199 if you're in Cohort 1. $399 after that.
The 2× jump for Cohort 2 is intentional and disclosed up front. Cohort 1 buyers get founding pricing in exchange for being the feedback layer that shapes Cohort 2. That's the deal — there's no hidden third price.
Founding Cohort
$199
$8.30 per live hour · $12 per week
- 16 live sessions · 24 hours total
- Weekly office hours
- Personal capstone PR review
- Demo Day + cohort certificate
- Private Discord access
- Lifetime recordings
- Founding Member badge for LinkedIn
- Direct influence on cohort 2's curriculum
Standard (Cohort 2+)
$399
$17 per live hour · $25 per week
- Same 16 live sessions, refined from cohort 1 feedback
- Weekly office hours
- Personal capstone PR review
- Demo Day + cohort certificate
- Private Discord access
- Lifetime recordings
- Possible add-ons: mock interview, resume review (TBD)
Opens after Cohort 1 graduates. Join the waitlist below to be first to hear.
For teams
Bringing a whole team?
Private cohorts mapped to your stack and your career ladder. The curriculum stays free and open — companies license the delivery.
Where this fits
The Junior track is the first rung.
The AI Infrastructure Curriculum is a ladder of role-based programs. Cohort 1 launches the Junior track. The rest are on the roadmap and prioritized based on what cohort feedback says you want next.
Cohort 1 · live
Junior AI Infra Engineer
This Academy. 16 weeks, code-along + capstone, Demo Day graduation.
Planned
Senior AI Infra Engineer
Production at scale: SLOs, oncall, capacity, multi-tenant infra, real platform work.
Planned
MLOps Engineer
Pipelines, experiment tracking, model registry, feature stores, retraining at cadence.
Planned
ML Platform Engineer
The internal platform layer: paved roads, self-service, GPU scheduling, cost levers.
Planned
AI Infrastructure Architect
System design at scale, technology selection, multi-region, security & compliance.
Specializations
Plus role-specific deep dives
Performance, Security, Principal-track. Likely as shorter focused programs after the main ladder is in motion.
You're not locked out of higher tracks by starting here. Each program is standalone — you can take Junior, then jump straight to Senior or MLOps when they open. Or skip Junior entirely if you're past that level and apply to a later track when it launches.
Who's teaching
The instructor
Instructor
Joshua Ferguson
Senior Systems Engineer · Phoenix, AZ. 15+ years building production infrastructure across enterprise cloud platforms, public sector, and AI-adjacent tooling.
Focus areas: multi-datacenter automation, container orchestration, AWS at enterprise scale, CI/CD, and — over the last two years — multi-agent systems for AI-assisted development.
I've spent the last decade and a half wiring up the infrastructure that real businesses run on — multi-datacenter automation, container orchestration, AWS at enterprise scale, the CI/CD that lets a team ship without paging someone every Friday night. This isn't material I read about and turned into slides. It's material I've shipped, broken, fixed, and shipped again.
The reason I'm teaching AI infrastructure now is that I've been living the transition from traditional infrastructure to AI-adjacent platform work for the last two years. I'm currently building APEX, a multi-agent orchestration platform on Anthropic's Claude Agent SDK — a "product team in a box" with specialized agents for planning, implementation, testing, and review. The cohort curriculum is shaped by both halves of that work: the fundamentals that don't go away, and the ML-specific gotchas that traditional infra training misses.
The Academy is small on purpose. You'll know every other attendee's name by week 3. You'll get personal line-by-line feedback on your capstone PR. You'll be expected to show up. In exchange, you get the thing I wish someone had built for me 12 years ago.
Frequently asked
Honest answers.
This is the Junior track — what if I'm past that level?
This program targets engineers becoming production-ready Junior AI Infra Engineers. If you've already shipped ML services in production, run on-call for ML systems, or designed platform infrastructure at scale, the Senior / Architect / MLOps tracks (on the roadmap) are where you'd fit. Apply to the waitlist for those and we'll let you know when they open.
If you're a strong backend engineer who hasn't shipped ML infrastructure yet — this is for you. Most of cohort 1 will be in that situation.
Do I need an ML background?
No. You need to be comfortable in Python and willing to learn. The cohort uses a small pre-trained model (DistilBERT) so we can focus on the infrastructure around it. We don't train models from scratch.
How much time per week, total?
~3.5 hours: the 90-minute live session, ~2 hours of homework, and optional office hours. Homework is scoped so a working adult can keep up over a weekend.
What if I miss a session?
Recordings are posted within 24 hours. The cohort starter repo is tagged at each
week's end (week-NN-end), so you can git fetch to where the
cohort landed and resume. Two-in-a-row absences get a friendly "you okay?" check from
the instructor.
What does Demo Day actually look like?
Week 16, second half. Each attendee who opts in gets 5-7 minutes to present their capstone: problem → architecture → one cool thing they built → metrics → what's next. Cohort + instructor Q&A after each. Recorded with your permission.
Opt-in. About half the cohort typically presents. Both presenters and watchers get the same certificate.
Will this get me a job?
The cohort makes you a stronger candidate. It does not guarantee an offer. What we can guarantee: you'll have a real production-shaped project you can demo, a vocabulary that matches what hiring managers ask about, and a portfolio repository that's interview-ready.
What hardware do I need?
A laptop with 8 GB+ of RAM and 30 GB free disk. macOS, Linux, or Windows + WSL. The cohort project runs locally up through week 12; weeks 13-14 use AWS free-tier (about $10-20 in costs if you don't tear down promptly).
Why $199 instead of $399?
You're the first cohort. You get founding pricing in exchange for being the feedback layer that shapes how cohort 2 is delivered. Cohort 1 testimonials and retrospective directly influence cohort 2's curriculum. That's worth $200 to us.
Refunds?
Full refund through the end of week 2 if it isn't a fit. After that, no refunds — the cohort is small and your seat blocks someone else.
When does Cohort 1 start?
Dates TBA. Join the waitlist below; you'll get the cohort dates and the payment link before the public launch.
Apply for Cohort 1
Tell me about you.
The cohort is capped at ~30 seats. I read every application personally. You'll hear back within 3 business days with cohort dates and payment link if accepted.