Anvil AI Academy

Programmes / Systems

Systems: agents and models

You will wire up an agent, give it a budget and logs, and learn to tell when a model needs tuning and when a better prompt is enough.

1agent running unattended, with a spend cap and a log you can read

A target we teach towards, not a guarantee.

Length
6 weeks
Format
Live online eveningsTue and Thu, 7:30 to 9:00 pm IST, plus one Saturday lab a week
Cohort size
20 seats
Fee
₹24,000Indicative. Assessment fees are separate.
See start datesFees and refunds

Who it is for

Engineers and researchers.

Strands covered

  • Strand 8, Agents and multi-agent systems
  • Strand 9, Models: tuning, training, serving

Week by week

Modules run from level 1 to level 3 across the strands above, and each follows the same seven steps. The last week is a full mock assessment and the defence of your assignment.

  1. Week 1

    • What an agent is; tools, MCP, function calling Strand 8, level 1
    • How LLMs are built: tokens, embeddings, attention, intuitively Strand 9, level 1

  2. Week 2

    • Build your first agent; guardrails and human-in-the-loop Strand 8, level 2
    • Datasets: collecting, cleaning, labelling, synthetic data Strand 9, level 2

  3. Week 3

    • Multi-agent orchestration: roles, briefs, handoffs, evidence Strand 8, level 3
    • Autonomous loops, scheduling, monitoring, kill switches Strand 8, level 3

  4. Week 4

    • Fine-tuning (LoRA, full); when it beats prompting and RAG Strand 9, level 3

  5. Week 5

    • Evaluation and benchmarks Strand 9, level 3

  6. Week 6

    • Full mock assessment, then the defence of your assignment before the cohort.
    • Book your live assessment with AIQ Institute.

What you will be able to do

SparkLevel 1

Strand 8, Agents and multi-agent systems

  • Explains agents, tools, MCP and function calling
  • Runs an off-the-shelf agent safely

Strand 9, Models: tuning, training, serving

  • Explains tokens, embeddings and attention intuitively
  • Reads a model card critically

TorchLevel 2

Strand 8, Agents and multi-agent systems

  • Builds a first agent with guardrails and a human in the loop
  • Logs what the agent did and why

Strand 9, Models: tuning, training, serving

  • Collects, cleans and labels datasets
  • Generates and audits synthetic data

BeamLevel 3

Strand 8, Agents and multi-agent systems

  • Orchestrates multi-agent systems with roles, briefs, handoffs and evidence
  • Runs autonomous loops with scheduling, monitoring and kill switches

Strand 9, Models: tuning, training, serving

  • Fine-tunes with LoRA or fully and shows when it beats prompting and retrieval
  • Serves a tuned model and evaluates it

Prepares for AIQ Institute levels 1 to 3

Prepares for AIQ Institute levels 1 to 3 in strands 8 and 9: Agents and multi-agent systems; Models: tuning, training, serving.

Assessment and certification are conducted by AIQ Institute, an independent standards body; assessment fees are paid to the institute directly.

See start datesTrain a team on this