BUILDYOUROWN AIDATAAGENT
Ready to move from dashboards to agents? In six guided one-hour sessions, build a locally-hosted data agent — your data, an AI API key I provide, and a Python harnessA harness is the wrapper that drives the model. For example: Claude Opus is the model; the Claude desktop app is the harness — it uses Opus to answer your questions by supplying the context, the tools, and the output format. you control. You'll also make it a public data agent on a dataset of your choice (or use one of mine).
- Build a real data agent
- Explore and optimise every layer (system prompt, context, eval)
- Learn the loop for using AI for real work
Who is this for?
- You're a data / analytics person who is ready to move from chat to code and get more out of AI
- You are a team lead / manager / HOD who wants a first-hand feel of what an analytics agent really is
- You learn best by doing, but get put off by technical details like API keys, service accounts, and harnesses
Who it's not for: anyone looking for a passive lecture, or someone who wants to become an AI expert in six hours.
Anyone already shipping agents in production every week.
Small Cohorts Only
These cohorts are meant for hands-on, interactive building — like a team working together. Cohort size is capped at 4 participants, and we share notes, wins, and frustrations along the way. Everyone is vetted for data competency before the cohort starts.
Most importantly, you get the time to get stuck on a problem and work your way through it with AI — which is central to actually using AI for work. And when you need help, I get you unstuck and moving to the next step.
Pick your cohort
Choose a cohort depending on your availability. Each cohort runs six one-hour sessions across a few days. Weekday cohorts are blocked for team bookings.
Tap a highlighted week for dates, seats & booking →
Seats are held on payment. Colours update as cohorts fill.
3 ways this moves you ahead on AI
It's one skill underneath
A harness, tools, context, evals — every agent is the same handful of parts. Build one small data agent and you can read (and build) the big ones.
Small is how you actually see it
Point an agent at your whole warehouse and you'll hit errors and think "agents aren't that good." Build a small one and you watch how it works — then take that to the data that matters.
Local first is the fastest way to learn
Running on your machine, with your key, you can see every call the agent makes. No black box — which is exactly what makes it click.
For team weeks, you choose what your team's agent should do.
Build your own AI agent in 6 sessions
Six guided one-hour sessions take you from an empty folder to a working data agent on your machine. Open any session to see what happens in it.
New to agentic analytics? Watch the 5-minute deck first →
- ›Install Python + Claude Code — I walk you through it live
- ›Connect to your data and drop in the AI API key (I provide it)
- ›Define the project: what your agent should answer, scoped so it's buildable
- you ship A ready environment — data connected, key in place, project defined
- ›Write your first agent harness in Python — the loop that talks to the AI
- ›Send a prompt, get a response, and see the request/response with your own eyes
- ›Understand what an "agent" actually is under the hood — no magic
- you ship A minimal Python agent you can run and talk to
- ›Give the agent tools — let it query your data, not just chat
- ›Ask a real question and watch it pick a tool, run it, and answer
- ›Fix what breaks with the agent's help, not Stack Overflow
- you ship An agent that answers a real question about your data
- ›Shape the plan → query → check → answer loop the agent follows
- ›Handle the messy cases: bad joins, empty results, follow-up questions
- ›Save your work in Git so nothing is lost
- you ship A reliable workflow, not a one-off lucky answer
- ›Write the context layer — the language of your business the agent reads
- ›Teach it your tables, metrics, and the gotchas every analyst knows
- ›Watch its answers get sharper as the context gets better
- you ship A context layer that makes the agent speak your data's language
- ›Write evals — a set of questions with known-good answers
- ›Run them to see, objectively, whether the agent actually works
- ›Learn the loop: eval → fix context → eval again
- you ship An evaluated agent — you know it works, and where it doesn't
- ›Wrap your agent in a simple Streamlit UI
- ›Deploy it to a public link you can send a friend
- ›Learn how to keep building after the cohort
- you finish with A shareable agent app — a real link, not just localhost
How your repo looks after the session Wondering what a repo is? In simple words, it's your work folder in Git. We cover it in the first session.
This loop is the learning.
Working with AI is all about mastering the loop — build, run, check, improve the context, run again.
Price
Note: Payment through UPI only for now. I will share the payment link once you fill up the form.
- A personal laptop (or a work laptop with admin rights to install apps)
- No API key needed for AI calls — I provide this
- No coding background needed — we set up your environment together in session 1
Not ready? Start with a free workshop
I run free build-along workshops on Luma — a low-stakes way to see how this works before you commit to a cohort.
See all workshops on Luma →Join now
If you are part of my WhatsApp community, DM me on WhatsApp.
Otherwise, fill the form below and I will DM you on WhatsApp/Email.
- I'll send you the payment link and confirm your seat on payment
- You get the setup slot + calendar invites
Questions? First check the FAQs, then DM me your question on WhatsApp.
Refunds
- 🔄 Full refund after Session 1. Come to the first session (setup + scoping). If you decide it's not for you, just message me after the session — full refund instantly, no questions asked.
- 🎥 Lifetime recordings. Every session is recorded and yours forever, so a missed session or a work emergency is never a loss.
- If you miss the whole cohort due to any emergency, you can simply take a seat in the next cohort or ask for a refund.
FAQs
Both — the agent you build is small enough for a complete beginner, and the pattern (harness, tools, context, evals) is worth building even if you're senior. If you've never coded, you're fine. Claude Code does most of the typing; you drive.
Not at all. That's exactly what session 1 is for — we set up your environment together and Claude Code does the heavy lifting. You'll read and run Python with the agent's help, not write it from scratch.
The cohort fee (₹6,999, or 50% off for community members), plus your own AI subscription for the build tools (e.g. Claude Code). I provide the API key for the agent's AI calls, so there's nothing extra to buy there. Running the agent locally is free.
I provide the key; we typically build on Claude. The pattern you learn — harness, tools, context, evals — is the same across models, so you're not locked to one.
About 6 hours total — six one-hour sessions, plus an optional session to put it online. Nothing required between sessions.
Sessions are recorded, so you can always catch up. We keep the cohort tight so you don't have to.
Yes — it's your repo, running on your machine. You keep it and can keep building on it. The optional Streamlit session puts it online too.
A working data agent, your own Git repo with the whole build, the experience of the agent build loop, and lifetime recordings. But really — you'll walk away understanding what an analytics agent actually is. You only get that by building one.
You start using this on your actual work — pointing agents at your dashboards, warehouse, or GitHub via MCP, writing and running queries, building semantic layers. The same skills apply in whatever you do next. We discuss this on the call while the agent builds.
No. Workshops are focused on small wins — like setting up a project or a simple automation. To really build an agent end to end, you have to spend focused time doing it. That's what this is.
No. That's a separate, much deeper 6-week program. This is the hands-on taster: six hours, build one small agent, feel how it works.
Fill up this form : https://tally.so/r/1AQvk4 (it takes 30 seconds).
Still have questions? DM me on WhatsApp.