AIUni
Design path

Learning Path Designer

A split view keeps the advisor and roadmap side by side. The chat stays embedded—no floating bubbles or distractions.

AI Learning Advisor

Embedded chat; grounded suggestions replace this placeholder once wired.

Prototype
Advisor
Welcome. Tell me where you are starting from, and I will shape a clear path with you.
You
I know Python and stats, but I am new to modern LLMs.
Advisor
Placeholder: When the backend is wired, this advisor will propose a staged plan with checkpoints and resources that match your background.
Placeholder UI; backend integration will deliver real guidance.

Personalized roadmap

A calm outline to track your progress.

Structured

Foundations

Start here

Clear math, programming, and data basics before touching heavier models.

  • Strengthen Python + notebooks for experiments
  • Probability, linear algebra, and calculus refreshers
  • Data literacy: cleaning, charting, and reasoning
  • Evaluate with baselines before adding complexity

Core ML

Supervised/unsupervised patterns, evaluation, and iteration discipline.

  • Modeling patterns: regression, classification, trees, ensembles
  • Cross-validation, leakage checks, and calibration
  • Feature pipelines and monitoring basics

NLP

Classical NLP groundwork to understand why LLMs behave as they do.

  • Tokenization, embeddings, and sequence models
  • Information retrieval and ranking fundamentals
  • Evaluation with human-in-the-loop checks

LLMs

Modern large language models with an emphasis on safety and reliability.

  • Prompt design with guardrails
  • Evaluation: correctness, safety, and user experience
  • Model selection trade-offs and latency budgeting

RAG

Retrieval-augmented generation with grounded context and observability.

  • Indexing strategies and chunking choices
  • Retrieval quality checks and feedback loops
  • Caching, fallbacks, and outage playbooks

Agents

Tool-use, planning, and failure containment for agentic systems.

  • When to use agents vs. simpler flows
  • Action orchestration, timeouts, and rollbacks
  • Human override and safety gates
Placeholder copy: Once the backend is connected, this roadmap will update live based on the advisor's reasoning and your feedback.