Video summary

AI Engineer Roadmap 2026: Become the top 1%

Main summary

Key takeaways

Educational

Main ideas / lessons conveyed

  • AI engineering is a high-paying, fast-growing field, with strong demand expected to continue rising through 2030.
  • A degree alone is not enough; many degree-holders struggle to get jobs due to a large AI skills/talent gap.
  • To become an AI engineer competitively, follow a step-by-step roadmap built on:
    • Foundational computer science skills (programming, DSA, APIs, cloud)
    • Data handling
    • Math basics
    • Classical ML + Deep Learning
    • Then optionally LLM-focused production skills to stand out
  • Avoid common career-killing mistakes: passive learning, learning in isolation, and applying too late.
  • Success is framed as a matter of action, discipline, and consistency, not just watching videos.

Roadmap to becoming an AI engineer (phases)

Phase 1 — Foundation (learn the basics first)

  1. Learn Python (core language for AI/ML)

    • Build small projects immediately, e.g.:
      • file organizer
      • calculator
      • webscraper
    • Use a coding assistant (recommended: Cursor or “Pilot GH”) to keep up in 2026.
    • Suggested schedule: 3–4 hours/day → learn Python in 3–4 weeks.

  2. Learn Git

    • Purpose: manage code safely and collaborate.
    • Suggested duration: ~1 week
    • Put every project on Git.

  3. Learn APIs

    • Understand basics → learn making API calls → build your own APIs
    • Tools/frameworks: FastAPI and Flask
    • Suggested duration: 2–4 weeks

  4. Learn Cloud platforms

    • Cloud essentials: GCP, AWS, Azure
    • Recommendation: start with AWS (most popular)
    • Suggested duration:
      • cloud basics
      • core cloud services
      • total: 3–4 weeks for real exposure

Phase 2 — Data handling skills (prepare and work with real data)

  1. Learn SQL / database query & manipulation

    • Data is commonly stored in relational systems and cloud warehouses.
    • Examples mentioned: MySQL, PostgresSQL, BigQuery, Snowflake
    • Suggested duration: 3–4 weeks

  2. Learn Pandas

    • Claim: 60–70% of AIML work is cleaning/exploring/transforming data.
    • Suggested duration: 2–3 weeks

  3. Learn NumPy

    • Focus: matrices, vectors, calculations, efficient number processing.
    • Suggested duration: 2–3 weeks

  4. Learn Data visualization

    • Tools mentioned: Matplotlib and Seaborn
    • Suggested duration: 1–2 weeks

Phase 3 — Math + Real Machine Learning (core modeling)

  1. Learn math basics (enough to understand how models learn)

    • Don’t go extremely deep; learn “just enough” to interpret ML.
    • Example interview target: explain gradient descent.
    • Suggested duration: 4–6 weeks

  2. Learn classical machine learning

    • Purpose: concepts behind applications like:
      • self-driving/vision-related ideas
      • fraud detection
      • recommendation systems
    • Interview focus: know which algorithm to use when
    • For every algorithm, learn:
      • what it is
      • when to use it
      • when not to use it
    • Tool: scikit-learn
    • Suggested duration: 5–6 weeks

  3. Learn deep learning

    • Focus on modern DL foundations used for:
      • image recognition
      • language models
      • speech
    • Mentioned topics: Transformers and Attention mechanisms (foundation of LLMs)
    • Library/framework mentioned: PyTorch (not TensorFlow)
    • Suggested duration: 4–5 weeks

  4. Build a first end-to-end portfolio project

    • Requirements:
      • predictive problem that’s realistic (e.g., fraud detection, house price, customer churn/search)
      • full pipeline from data collection → training → evaluation
      • include a proper README
    • Post on GitHub.

Additional guidance in this phase

  • For freshers: classical ML + deep learning are enough.
  • Avoid attempting to learn “everything”; learn the right things at the right time.

Phase 4 — “Money face” / LLM production skills (optional, but job-boosting)

(Framed as not mandatory for freshers, but highly differentiating—targeting skills companies pay ~₹15–20 LPA for.)

  1. Learn working with LLMs programmatically

    • Rationale: “almost every product is made on LLMs”
    • Suggested duration: 1–2 weeks

  2. Learn RAG (Retrieval-Augmented Generation)

    • Claim: LLMs don’t inherently hold your personal data; RAG enables answering from your documents.
    • Suggested duration: 2–3 weeks
    • Goal: create a working RAG system

  3. Learn LangChain and LangGraph

    • Purpose: build RAGs and agents at scale, production-level structures.
    • Suggested duration: 1–2 weeks

  4. Build two key portfolio projects (job-getters)

    • Project 1: RAG chatbot that answers questions from a document corpus
    • Project 2: AI agent that automates a useful task
    • Resume requirement: include a live URL
    • Encouragement: stand out by sharing a live AI chatbot.

Crucial mistakes to avoid (detailed)

Mistake #1: Passive learning (“tutorial hell”)

  • Problem pattern:
    • watch 10–20 hour courses
    • delay building until later
  • Why it fails:
    • you “forget 70%” by course end
    • you get an illusion of learning, then blank out when you try to build
  • Fix (recommended practice ratio + schedule):
    • Practice 2–3x more than you learn
    • Example for 4 hours/day:
      • 1 hour concept learning (videos/docs)
      • 2 hours hands-on building using tools immediately
    • Integrate AI-assisted coding from day one (Cursor / “Pilot GH”).
    • Also practice DSA:
      • DSA is “almost mandatory for placement”
      • solve 1–2 problems daily

Mistake #2: Learning in isolation

  • Problem pattern:
    • struggling alone for hours
  • Why it fails:
    • sticking points get solved faster with others
    • you lose motivation when solo
  • Fix:
    • find buddies/partners:
      • Lindin/LinkedIn
      • Discord
      • communities on Reddit / Kaggle
    • do mock interviews with each other
    • gain clarity by discussing, explaining, and teaching

Mistake #3: Applying too late

  • Problem pattern:
    • “I will learn everything first, then apply”
  • Fix:
    • start applying as soon as Phase 4 is over
    • apply when you have:
      • Python
      • basic ML
      • at least a project
    • internship/work experience beats multiple personal projects
    • every month delayed = missing real-world experience

Motivation / framing at the end

  • The speaker claims most people won’t act after the video:
    • 90% do nothing
    • 9% start but quit after motivation drops
    • only 1% persist until they get a job (~₹15 LPA)
  • The decision is whether the viewer will take consistent action.

Speakers / sources featured

  • Speaker: Akbar (Senior Software Engineer in the UK for JP Morgan “JP mortgages” mentioned; described as working at a world #1 investment bank and #1 in AI adoption)

  • Sources/tools/frameworks named (learning tools referenced):

    • Cursor / “Pilot GH” (coding assistants)
    • Python
    • Git
    • APIs (and frameworks FastAPI, Flask)
    • AWS (plus GCP, Azure mentioned)
    • SQL dialects/engines: MySQL, PostgreSQL, BigQuery, Snowflake
    • Pandas
    • NumPy
    • Matplotlib, Seaborn
    • scikit-learn
    • PyTorch (referred to as “Pitos” due to subtitle errors)
    • LLMs
    • RAG
    • LangChain, LangGraph
    • DSA (data structures & algorithms)
    • Platforms/communities mentioned: LinkedIn, Discord, Reddit, Kaggle
    • Learning resource mention: GATE (referenced concept in the intro)

Original video