About BrownMind

Meet the founder

Apurva Khandelwal

Founder & AI Systems Architect at BrownMind

I design and build AI software for businesses. I founded BrownMind in 2022, and we build custom AI software for sales teams.

My focus is software that makes the decisions an experienced person makes, and that a team can test and check. Examples are checking leads, reviewing sales calls and helping an owner manage daily work.

Read about my areas of expertise
Apurva Khandelwal, founder of BrownMind
Based in Jaipur, India

Areas of expertise

How I turn business decisions into supervised AI software

  • Turning expert decisions into testable software

    I model how an experienced person decides, then test the software against their decisions. Lead research and sales-call review are examples: the rules and evidence matter as much as the answer.

  • Keeping each assistant's data and actions separate

    I design AI assistants with their own data access, allowed tools and approval rules. Each business action needs a record that a person can check. Shared information should not expose another customer's private records.

  • Preparing a business for AI-assisted work

    I focus on the process an assistant needs to follow: trusted information, decision rules and what to do when something goes wrong. People need to know what the software may do and when it must ask for help.

Experience

From business and marketing work to AI engineering

My experience spans running a business, building software and working in AI engineering teams. I founded AG Advisers in 2016 and BrownMind in 2022.

That background shapes the questions I ask before building: how does the team decide today, what information does it trust, and who checks a mistake?

Read the BrownMind story
  1. March 2016 to March 2022

    Founder at AG Advisers

    My earlier business work included growth and performance marketing, alongside websites and applications for local businesses.

  2. March 2022 to present

    Founded BrownMind

    BrownMind now builds custom AI software for sales teams and agencies.

  3. October 2025 to October 2026

    Senior AI Engineer at Ante Digital (Canada)

    A previous engineering role. Salesight was built with Ante Digital, and its case study below gives the exact credit.

  4. October 2026 to present

    Chief Distribution Engineer at Spektra (Sweden)

    My current role, alongside running BrownMind.

Selected work

Projects and my role in them

Each case study says what the software does and who built it. The credit on each project states my role. You can also visit Salesight.ai.

  • Salesight's public overview and call-review screens.

    Sales coaching

    Salesight scores every sales call against your own script

    Before
    A manager listens to a few calls a week by hand.
    After
    Every recorded call is scored from 0 to 100 against the team's own script, with a view for each rep.

    Built with Ante Digital (Canada), with our founder as lead developer.

  • Three property messages become one reply, using the stated budget, timing and cash payment details.

    Real estate

    One WhatsApp reply to every property lead, saved in the sales system

    Before
    Leads kept in a spreadsheet, and brochures sent by hand.
    After
    The assistant answers grouped messages, saves details in Close, the team's sales system, and offers eligible buyers a viewing link.

    Designed and built by our founder, as lead developer for a partner agency.

  • Five company checks: two pass, two fail and one needs review. One email draft is ready and one is blocked. Sending stays off.

    B2B outbound

    Check each company's fit before paying for contact research

    Before
    One prospect researched at a time, with decisions scattered across spreadsheets.
    After
    Companies are checked for fit first, and a person reviews each sourced email draft. Email sending was not switched on.

    Reviewed, hardened and operated by our founder as part of a partner agency's team, working with the client's technical founder.

How I build

Test decisions before allowing customer-facing actions

A convincing demonstration is not enough to give software permission to act. My approach is to check its decisions, limit what it can do and keep a person responsible for higher-risk actions.

My tools include TypeScript, Node.js, Python and Astro. I use them to connect business tools and build the screens people use.

  • Compare answers with expert decisions

    Use decisions from the team's work to test the system before launch. Check missing information and exceptions, not only straightforward cases.

  • Keep evidence and approvals beside the action

    Make the supporting information checkable. Require approval for customer-facing actions until the system has earned that trust. Each project still needs its own permissions and checks.

  • Control costs and improve after launch

    Set a cost cap for each run and review failures after launch. Improve against the team's agreed measures, rather than treating the first working version as finished.

Writing

Technical ideas I write about

In my featured LinkedIn posts, I explain a design that gives each broker a separate AI assistant. The assistants share one set of tools, and each tool checks permission before it acts. The design keeps conversations apart from data access and records each business action.

I also write about making a company ready for AI: documenting decisions, resolving conflicting information and defining when a person must step in. These posts describe my approach. They do not say that every project already follows it.

Read my featured posts on LinkedIn

Profiles and contact

Find my work and get in touch

This profile draws on BrownMind's company story, the linked case studies, my LinkedIn experience and writing, and my public GitHub profile. The case studies give project-specific credits and limits; my writing explains the approach I am developing.