DV-Gym | Stage 1: AI-Era Design Verification Foundations

Stage 1: AI-Era Design Verification Foundations

Bridging Modern AI Capabilities with Hardware Verification Workflows

Course: DV-Gym — Stage 1 Foundations
Instructor: Ramdas Mozhikunnath
Audience: Verification Engineers - Juniors, Seniors, Leads and anyone else curious

www.verificationexcellence.in
DV-Gym | Stage 1: AI-Era Design Verification Foundations

Course Scope & Mindset

🎯 What This Course Is

  • Practical Mental Models: Understanding how LLMs and agents actually work under the hood.
  • DV Application Focus: Direct mapping to Design Verification flows and skills
  • Practical Engineering: Knowing when to use AI, how to steer it, and where it fails in silicon workflows.

🚫 What This Course Is Not

  • Not a Pure ML/Math Course: No gradient descent derivations, backprop proofs, or tensor calculus.
  • Not AI Hype: No magical "push-button autonomous tape-out" claims.
  • Not a Syntax Replacement: EDA simulators, formal tools, and human judgment remain the final authority.

Guiding Principle: AI does not replace verification expertise—it accelerates the engineer who understands its mechanics and boundaries.

www.verificationexcellence.in
DV-Gym | Stage 1: AI-Era Design Verification Foundations

Prerequisites & Audience

🛠️ Verification Background (Assumed)

  • Familiarity with Design Verification basics - SV,UVM basics, Simulations, Regressions, Coverage.
  • Basic understanding of simulation cycles, compile errors, and regression triage.
  • Working knowledge of Linux terminal commands, Git, and EDA scripting glue.

🤖 AI Background (None Needed)

  • No previous experience with AI/ML development required.
  • Curiosity to move beyond standard chat windows to closed-loop engineering workflows.
  • Ready to treat all model outputs as testable hypotheses rather than verified truth.
www.verificationexcellence.in
DV-Gym | Stage 1: AI-Era Design Verification Foundations

Stage 1 Agenda: Lectures 1 & 2

📘 Lecture 1: Foundations of LLMs

  • What are LLMs, How they work?, Tokenization, context windows.
  • Prompting, Context - Good and bad practices.
  • Practical examples.

🤖 Lecture 2: Agentic AI & Closed Loops

  • Moving from passive chatbots to autonomous agents.
  • How tool calling works with EDA simulators.
  • Guardrails, agent traps, and Human-in-the-Loop gates.

 

www.verificationexcellence.in
DV-Gym | Stage 1: AI-Era Design Verification Foundations

Stage 1 Agenda: Lectures 3 & 4

📋 Lecture 3: Skills & The MCP Standard

  • Packaging reusable playbooks (uvm-hang-triage, sva-from-spec).
  • Differentiating Skills, Tools, and Project Rules.
  • Model Context Protocol (MCP) as the universal tool interface.

🧩 Lecture 4: Orchestrating the DV Flow

  • Multi-agent coordination (DV Lead mental model).
  • The Execution Harness vs. The LLM Brain.
  • RAG for dense protocol specs and IP safety guardrails.

 

www.verificationexcellence.in
DV-Gym | Stage 1: AI-Era Design Verification Foundations

Essential References: LLM & Prompt Foundations

  • Prompt Engineering Guide (DAIR.AI): Comprehensive compendium on zero-shot, few-shot, and reasoning techniques (promptingguide.ai).
  • Anthropic / OpenAI Interactive Cookbooks: Best practices for function calling, context structuring, and tool execution.

 

www.verificationexcellence.in
DV-Gym | Stage 1: AI-Era Design Verification Foundations

Essential References: Architecture, Standards & Verification

  • Model Context Protocol (MCP) Specification: The open standard for connecting AI clients to data sources (modelcontextprotocol.io).
  • DeepLearning.AI Short Courses: Practical bite-sized modules on building agentic workflows and tool-calling harnesses.
  • Verification Excellence: Reference articles on SystemVerilog, Assertions, and AI in DV (verificationexcellence.in).

Next Up: Let's jump into Lecture 1: Introduction to LLMs & Prompting Foundations.

www.verificationexcellence.in