Agentic AI with Hermes Agent
A free, hands-on 2-week intensive. No prerequisites — just curiosity. Everyone welcome.
Schedule: Mon–Wed–Fri, 6–7pm MST (7 sessions), Fri Sep 25 – Fri Oct 9, 2026. Day 1 is a talk-only kickoff; hands-on labs run Day 2 through Day 7.
Time zones: 6pm MST (Calgary) · 5pm PST · 7pm EST · 7:30pm IST
By the end, you'll have deployed a working Hermes agent — your own digital agent, working and learning for you 24x7. Three months ago, following Sanjay Jayaram's vision, we purchased a DGX Spark Supercomputerwith a Blackwell GPU and 128 GB of memory, powered on here in Calgary — an NVIDIA® GB10 Grace Superchip with 6,144 CUDA cores capable of 1 Petaflop FP4 compute, enough to run up to a 200-billion-parameter LLM or several 30-billion-parameter models at once. Sanjay also picked up a powerful Apple Mac Mini for local AI research after interning at Apple in Cupertino. We'll be setting up and demoing our AI research on both machines live during this course, and Sanjay will join some evenings to share what he's learned.
Day 1: Intro to LLMs and Hermes
Fri, Sep 25, 2026
Talk only, no lab — the big picture on LLMs, what agentic AI means, and why Hermes Agent is our tool for the course.
Day 2: Hermes Fundamentals
Mon, Sep 28, 2026
What Hermes Agent is, how it talks to any OpenAI-compatible endpoint, and getting it running against our local LLMs on the DGX Spark and Mac Mini.
Day 3: LLM Routing with Hermes
Wed, Sep 30, 2026
Pointing Hermes at multiple local models and routing requests between them based on task.
Day 4: RAG with Hermes
Fri, Oct 2, 2026
Grounding Hermes Agent in real data instead of letting it guess.
Day 5: MCPs for Hermes
Mon, Oct 5, 2026
Connecting Hermes Agent to external systems and tools over MCP.
Day 6: Tool Calling & Agent Loops
Wed, Oct 7, 2026
How Hermes decides to call a tool, reads the result, and loops until the task is actually done.
Day 7: Deploy a Production Hermes Agent
Fri, Oct 9, 2026
Capstone: ship a Hermes agent that keeps running and working for you 24x7.
Day 8: Build a Chatbot
Bonus session
Bonus: build a safe public AI chatbot with a web page, a small server that guards the secret key, and a model that writes the answers.
Hermes Agent Capability Map
Everything Hermes Agent can do, on one page — tools, memory, skills, subagents, scheduling, MCP, providers, messaging, backends, and more. Click the image to open it full size.
Inside Hermes Agent
How it all fits together: one agent loop — receive, think, act, observe — wrapped by entry points, swappable model providers, tools and skills, and memory, with each part tagged by the course day that covers it. Click the image to open it full size.
