About this roadmap

A personal learning tracker built for Shubham Arora — backend engineer with 5+ years across 4 companies, pivoting into building and deploying LLM systems. 5–8 hrs/week, ~14–20 calendar weeks to portfolio-ready.

What's in here

Roadmap

Full 7-level curriculum with skills, resources, and progress tracking

Project Log

Track projects you've shipped at each level

Reading List

15 curated papers and blogs — the foundational mental models

Interview Prep

System design prompts and behavioral questions for LLM roles

The honest caveat

Most “2026 roadmap” sources are published by course-sellers; the urgency framing is partly marketing. The skill demand is real and well-corroborated across independent job postings — the panic is not.

Alternate courses with building. Never stack courses without shipping. Portfolio > certificates, especially given an existing engineering track record.

The edge

The 2026 market wants “can you ship AI to production reliably,” not “can you train a model.” Ops, eval, and deployment skills transfer directly from existing backend experience and are a differentiator vs pure-ML candidates.

Target roles (demand-ranked)

AI / LLM Product Engineer — GenAI Engineer, Forward Deployed Engineer. Best fit.
LLMOps / MLOps Engineer — Close second; leans on existing infra skills.
LLM Engineer (RAG-focused)
AI Solutions Architect — Later-career.

Geography

India roles (incl. remote-for-US/EU) and relocation/sponsorship markets (Germany, Netherlands, Canada, UK, Singapore, Ireland). The roadmap is geography-agnostic at the skill level; cloud-platform choice in Level 4 is the only geo-sensitive fork.

Stack

Next.js 16 · TypeScript · Tailwind v4 · shadcn/ui (base-nova / @base-ui/react) · react-markdown · localStorage for all persistence · no backend, no auth.