Learning Paths
Applies to: every Lieber engineer and presales engineer · Typical shape: one path per role, ~3–6 months part-time for the required items, then recommended items as projects demand
Each path lists required items (the baseline we expect before you lead that kind of work) and recommended items (depth for specialists). Effort figures are approximate and assume part-time study alongside project work. Prices, exam codes, and course line-ups were checked in October 2026; vendors update them often, so confirm on the linked page before booking.
Common core (all roles)
| Resource | Type | Effort | Status | Link |
|---|---|---|---|---|
| How We Build (this site) | Internal handbook | 2 h | Required | How We Build |
| Local AI Basics → Inference Engineering → Local RAG Lab | Internal pages + lab | 1–2 days | Required | Local AI Basics, Lab |
| Open Model Landscape + Local AI Best Practices | Internal pages | 2 h | Required | Model Landscape, Best Practices |
| Hugging Face LLM Course (chapters 1–4) | Free course | 10–15 h | Required | huggingface.co/learn/llm-course |
| AI Engineering — Chip Huyen (O'Reilly) | Book | 20–30 h | Required | chiphuyen/aie-book |
| AI-Native Engineering → Lab: Agent-Ready Repo | Internal pages + lab | 1 day | Required before your first agent-assisted client engagement | AI-Native Engineering, Lab |
| Kubernetes and Cloud Native Associate (KCNA) | Certification (multiple choice, US$250) | 20–30 h | Recommended | CNCF certifications |
AI engineer
Builds RAG systems, agents, evals, and fine-tunes. Services: GenAI PoC, RAG Systems, Agentic AI Platforms, Fine-Tuning, AI Productification.
| # | Resource | Type | Effort | Status | Link |
|---|---|---|---|---|---|
| 1 | Hugging Face LLM Course (all 12 chapters: transformers, fine-tuning, reasoning models) | Free course (no certificate yet) | 40–60 h | Required | LLM Course |
| 2 | Hugging Face Agents Course (smolagents, LangGraph, LlamaIndex; free certificate) | Free course | 30–40 h | Required | Agents Course |
| 3 | Hugging Face MCP Course (built with Anthropic; certificate) | Free course | 15–20 h | Required | MCP Course |
| 4 | AI Engineering — Chip Huyen, focus on ch. 3–4 (evaluation), 6 (RAG and agents), 7–8 (fine-tuning, data) | Book | (core) | Required | Book resources |
| 5 | AWS Certified Generative AI Developer – Professional (AIP-C01) | Certification (US$300, Professional) | 60–80 h | Required for cloud-AI projects | Exam guide |
| 6 | Hugging Face Context Course (skills, MCP, plugins, subagents, agent harness) | Free course | 10–15 h | Recommended | Context Course |
| 7 | Hugging Face smol course (post-training small models) | Free course | 15–20 h | Recommended | smol course |
| 8 | Andrej Karpathy — Neural Networks: Zero to Hero (micrograd → GPT-2) | Video series + notebooks | 30–40 h | Recommended | nn-zero-to-hero |
| 9 | fast.ai — Practical Deep Learning for Coders | Free course | 60–80 h | Recommended | fast.ai |
| 10 | AWS Certified Machine Learning Engineer – Associate (MLA-C02, beta since 29 Sept 2026; adds GenAI and agentic AI) | Certification | 40–60 h | Recommended | MLA-C02 update |
| 11 | OWASP Top 10 for LLM Applications 2025 + Agentic Top 10 2026 | Reading | 4 h | Required before shipping agents | OWASP GenAI |
| 12 | Hugging Face Open-Source AI Cookbook | Recipes | as needed | Recommended | Cookbook |
| 13 | Claude Certified Architect – Foundations (CCAR-F; 60 Q, 120 min, scenario-based) | Certification (US$125, partner programme) | 30–40 h | Recommended; map of domains to pages in AI-Native Engineering | CCA – Foundations |
Milestone: build the Local RAG Lab end to end, then extend it with a tool-calling agent behind an MCP server and an eval suite in CI.
Platform / cloud engineer
Builds landing zones, Kubernetes platforms, GPU clusters, and migrations on STACKIT, AWS, OpenStack, and Cloudflare. Services: Sovereign Cloud on STACKIT, Hybrid Cloud / OpenStack, Cloud Migration, Cloud-native Development, Private & Sovereign AI.
| # | Resource | Type | Effort | Status | Link |
|---|---|---|---|---|---|
| 1 | Certified Kubernetes Administrator (CKA) | Certification (performance-based, US$445) | 60–80 h | Required | CNCF certifications |
| 2 | STACKIT University — Certified STACKIT Cloud Engineer (free; tracks A–D from fundamentals to advanced architecture) | Vendor training + certificate | 30–50 h | Required | STACKIT University · Certification tracks |
| 3 | AWS Certified Solutions Architect – Associate | Certification | 60–80 h | Required for AWS projects | AWS Certification |
| 4 | Certified Kubernetes Security Specialist (CKS; requires CKA) | Certification (US$445) | 50–70 h | Required for sovereign/regulated projects | CNCF certifications |
| 5 | Certified OpenStack Administrator (COA) — OpenInfra Foundation, 3 h online, US$400, no retakes | Certification | 50–70 h | Required for OpenStack projects | openstack.org/coa |
| 6 | Cloudflare Learning Paths (Workers, Developer Platform, Workflows, Holistic AI Security) | Free vendor paths | 10–20 h | Recommended | Cloudflare Learning Paths |
| 7 | Canonical Academy — Canonical Qualified SysAdmin (4 browser-based exams: Linux Terminal, Ubuntu Desktop, Ubuntu Server, DevOps Principles) | Certification track | 30–40 h | Recommended | Exam guide |
| 8 | Certified Kubernetes Application Developer (CKAD) | Certification | 40–60 h | Recommended | CNCF certifications |
| 9 | Certified Cloud Native Platform Engineer (CNPE) / Associate (CNPA) | Certification | 50–70 h | Recommended for platform leads | Kubestronaut program |
| 10 | Prometheus Certified Associate (PCA) / OpenTelemetry Certified Associate (OTCA) | Certification | 20–30 h each | Recommended | CNCF certifications |
| 11 | NVIDIA GPU Operator + DCGM-Exporter docs | Docs | 6–8 h | Required before GPU clusters | DCGM-Exporter |
AWS Certified Machine Learning – Specialty retired on 31 March 2026. MLA-C01 was last offered in English on 28 September 2026; MLA-C02 runs in beta from 29 September 2026. Credentials earned on retired exams stay valid until their normal expiry.
Inference engineer
Sizes, serves, quantizes, and operates models on GPUs. Services: Private & Sovereign AI, Cloud AI Platforms. Internal pages: Inference Engineering, Serving at Scale.
| # | Resource | Type | Effort | Status | Link |
|---|---|---|---|---|---|
| 1 | DeepLearning.AI — Fast & Efficient LLM Inference with vLLM (Red Hat; LLM Compressor, vLLM, GuideLLM, lm-eval) | Free short course (~1.5 h, 9 lessons, 3 labs) | 4–6 h | Required | vLLM blog announcement |
| 2 | Inference Engineering — Philip Kiely (Baseten Books, 2026): models, hardware, engines, techniques, production | Book (free online chapters) | 15–20 h | Required | baseten.co/inference-engineering/book |
| 3 | vLLM paper — Efficient Memory Management for LLM Serving with PagedAttention (SOSP 2023) | Paper | 3 h | Required | arXiv 2309.06180 |
| 4 | vLLM docs: structured outputs, metrics, sleep mode; vLLM blog | Docs | 8–10 h | Required | Structured outputs · Metrics · vLLM blog |
| 5 | Stanford CS336 — Language Modeling from Scratch (Spring 2026; GPUs, kernels, Triton, parallelism, inference, scaling laws) | University course (free videos + assignments) | 100+ h | Required lectures on GPUs, kernels, parallelism, inference; assignments recommended | Lecture materials |
| 6 | FlashAttention-2 | Paper | 2 h | Required | arXiv 2307.08691 |
| 7 | FlashAttention-3 (Hopper: asynchrony, FP8) | Paper | 2 h | Recommended | arXiv 2407.08608 |
| 8 | FlashAttention-4 (Blackwell: pipelining co-design, CuTe-DSL) | Paper | 2 h | Recommended | arXiv 2603.05451 |
| 9 | AI Engineering — Chip Huyen, ch. 9 (inference optimization) | Book chapter | 4 h | Required | Book resources |
| 10 | Karpathy — Zero to Hero (for the model internals behind KV cache and attention) | Video series | 30–40 h | Recommended | nn-zero-to-hero |
| 11 | CKA (to run engines on Kubernetes) | Certification | 60–80 h | Required | CNCF certifications |
| 12 | llguidance / XGrammar (constrained decoding internals) | Repo + paper | 3 h | Recommended | llguidance · XGrammar-2 |
Milestone: serve one model from the landscape on vLLM in FP8 and on llama.cpp in Q4_K_M, benchmark both with GuideLLM at three concurrency levels, and publish TTFT, inter-token latency, tokens/s, and € per 1M tokens with Grafana dashboards.
Presales engineer
Explains architectures, sizes PoCs, and answers technical questions during scoping. Delivery teams own implementation; presales needs enough depth to scope correctly and set realistic expectations. Services: AI Assessment, Cloud & Migration Assessment, GenAI PoC.
| # | Resource | Type | Effort | Status | Link |
|---|---|---|---|---|---|
| 1 | AWS Certified AI Practitioner (AIF-C01; foundational, US$100, 90 min) | Certification | 20–30 h | Required | AWS Certification |
| 2 | AI Engineering — Chip Huyen, ch. 1 (building AI apps), 3 (evaluation), 10 (architecture) | Book chapters | 10 h | Required | Book resources |
| 3 | STACKIT University — fundamentals track | Free vendor training | 8–12 h | Required | STACKIT University |
| 4 | KCNA | Certification | 20–30 h | Required | CNCF certifications |
| 5 | Open Model Landscape (licences, tiers) + Best Practices (EU AI Act section) | Internal pages | 3 h | Required | Model Landscape · Best Practices |
| 6 | Hugging Face LLM Course chapters 1–2 | Free course | 5–8 h | Required | LLM Course |
| 7 | AWS Certified AI Business Strategist (beta) | Certification | 15–20 h | Recommended | AWS Certification |
| 8 | Commission guidelines for GPAI providers (EU AI Act) | Official guidance | 3 h | Recommended | GPAI guidelines |
| 9 | Cloudflare Learning Paths — Holistic AI Security | Free vendor path | 3–4 h | Recommended | Cloudflare Learning Paths |
| 10 | DeepLearning.AI short courses (pick 2–3 relevant to current pipeline) | Free short courses | 2–4 h each | Recommended | learn.deeplearning.ai |
Certification overview (as of 2026-10)
| Certification | Body | Format | Price | Notes |
|---|---|---|---|---|
| KCNA | CNCF / Linux Foundation | Multiple choice, online, proctored | US$250 (includes one free retake) | Part of the CNCF CARE maintenance programme; check renewal rules on the CNCF page |
| CKA / CKAD | CNCF / Linux Foundation | Performance-based | US$445 | Core Kubernetes |
| CKS | CNCF / Linux Foundation | Performance-based | US$445 | Requires CKA |
| CNPA / CNPE | CNCF / Linux Foundation | MC / performance-based | US$250 / US$445 | Platform engineering |
| AWS AI Practitioner (AIF-C01) | AWS | 65 Q, 90 min | US$100 | Foundational |
| AWS ML Engineer – Associate (MLA-C02) | AWS | Beta: 85 Q, 170 min | US$75 beta | Replaces MLA-C01 |
| AWS GenAI Developer – Professional (AIP-C01) | AWS | 65 scored Q | US$300 | RAG, agents, governance |
| Claude Certified Architect – Foundations (CCAR-F) | Anthropic (Claude Partner Network) | 60 Q, 120 min, proctored, pass 720/1000 | US$125 | Valid 12 months; domains: agentic architecture, tool design and MCP, Claude Code configuration, prompt engineering, context management |
| Certified OpenStack Administrator | OpenInfra Foundation | Performance-based, 3 h | US$400 | No retakes |
| Certified STACKIT Cloud Engineer | STACKIT University | Program + certificate | Free | Sovereign cloud baseline |
| Canonical Qualified SysAdmin | Canonical Academy | 4 browser exams | Not stated on the exam guide; check the Academy shop | Four exams: Linux Terminal, Ubuntu Desktop, Ubuntu Server, DevOps Principles |
How we track progress
- Agree your path with your tech lead and review progress together each quarter.
- Required items for a role are expected before you lead a project of that type; pair with a certified colleague until then.
- When you find a newer course, a retired exam, or a broken link, update this page and its sources in the same change.
Sources
- Hugging Face: Learn index · Agents Course · MCP Course · MCP Course repo · Hugging Face courses review 2026
- DeepLearning.AI and vLLM: vLLM course announcement · Red Hat course announcement · DeepLearning.AI platform
- University and self-study: Stanford CS336 lectures · Karpathy nn-zero-to-hero · fast.ai
- Books: AI Engineering resources · AI Engineering on Amazon · Inference Engineering book
- Papers: vLLM / PagedAttention · FlashAttention-2 · FlashAttention-3 · FlashAttention-4 · XGrammar-2
- Kubernetes / CNCF: CNCF certifications · Kubestronaut program · Golden Kubestronaut bundle · Kubernetes certifications 2026 (prices, formats)
- AWS: AWS Certification · AIP-C01 exam guide · MLA-C02 update · ML Specialty retirement · AWS AI certifications 2026 overview
- STACKIT: STACKIT University · STACKIT certification tracks
- OpenStack: Certified OpenStack Administrator
- Cloudflare: Cloudflare Learning Paths
- Canonical: Canonical Academy exam guide · Introducing Canonical Academy · DevOps Principles beta
- Anthropic: Claude Certified Architect – Foundations · Certification FAQ · Claude Partner Network announcement
- Security / regulation: OWASP GenAI resources · EU GPAI guidelines