Agent Force Optimization R&D
One agent. One hypothesis. One goal. At a Time.
I build Agentforce agents on Salesforce and ship them into real workflows — then publish the architecture, the failures, and the numbers.
Live now
Portfolio
of Agents
MyAiNativeTown · Agent Learning & Building Program
Cohort 01 · applications open
A 90-day program where freshers and working professionals learn to build AI agents by building them, starting with one real agent for Bhiwani and Hisar. The program runs its own agent, which answers questions about it and writes every log entry in public.
Ph.D.-Style Thesis: Agent God Complexes
Reimagining the Hiring Industry in the Age of AI.
3MoMLife — Discipline Agent
Salesforce Agentforce agent built on the 3MistakesOfMyLife.in discipline programme.
Hermes — Million Dollar Consistency Partner
Private agent — not yet public
A five-agent pipeline that turns daily working sessions into compounding output.
Next-generation builders
Why would someone building agents for global SIs start in Bhiwani?
I was born here and grew up here. It’s where I learned everything first, so it’s where I’ve chosen to build the team I’ll work with for the next decade. The manifesto came before the curriculum.
MyAiNativeTown is how that team begins: an agent learning and building program. Cohort 01 learns by shipping one real agent for two towns in Haryana, live, on alternate days. The program has its own agent too. It answers questions about the program and writes the log entries, so every hypothesis, action and wrong answer stays readable, including by you.
Their fee is due only when someone hires them. So I’m paid only if they’re worth hiring.
Some of them will build with me. Some will build with you. Whose log would you want to read first?
How it works
Exploratory Sprint
→ Measured Experiment
Four weeks from first call to a result you can defend in a board meeting. No retainer, no discovery theatre.
Map the system before touching it.
I trace your funnel end to end and find where work leaks. You leave with a written hypothesis, the assumptions it rests on, and the boundaries we agree not to cross.
The number gets agreed before the first line of code.
We build against a single measurable outcome — pipeline influenced — and the guardrails are documented up front, not patched in after something breaks.
It moved the number, or it did not.
Both are results. Most agent pilots never reach this point because nobody agreed what success looked like on day one.
About
Hari Om Vashishtha
I build agents on Salesforce and ship them into real workflows — then publish the architecture, the failures, and the numbers. Everything here has been run on my own operations before it is offered to yours.
- Salesforce
- Agentforce agents built and deployed on live orgs.
- Method
- Hypothesis, assumptions, boundaries — written before code.
- Measure
- Pipeline influenced. One number, agreed on day one.
Hire me
Tell me what’s breaking.
Not a newsletter signup. This goes straight into a Salesforce Lead record and I read every one myself.
- 1You describe the system that isn’t behaving. Be specific — vague inputs get vague replies.
- 2I reply within two working days with what I’d want to look at first, whether or not there is an engagement in it.
- 3If it is a fit, we scope the 5-day Exploratory Sprint. If it is not, I will say so plainly.
This form is a live Salesforce Web-to-Lead endpoint running on my own org — captcha, field mapping and lead routing included. The site is the first system I am asking you to judge.