Founding PM | craft, ownership, business

BITS Pilani (2014) | Postgrad ML Fellow, Plaksha (2021)

Business-oriented product builder with 10+ years of diverse work-ex - software, design, product roles.

Last role - Founding PM building an AI recruiter at Expertia AI (pre-Series A), growing the platform from 2M to 30M candidates over 2 years, led enterprise implementation for 500+ recruiters at the Reliance group, winning a successful renewal.

Formerly, Senior PM leading the learning outcomes data product for enterprise programs at upGrad's Harappa Education, self-taught software dev at Auroville, and designer at Mu-Sigma.

Featured

  • Building Ek-ai as a side venture - a proactive morning brief system that tells Indian business operators (WhatsApp-heavy) the three things that matter today like an annoyingly attentive EA.
  • A weekly AI coach that diagnoses why I get stuck, not just that I do. Open-source skill.
  • sid's pod craft - Mar 2023 AI episode when LLMs were just starting to have a moment
  • Spent 5 transformative years in Auroville after BITS.

Selected impact

  • Scaled from 2M to 30M candidates in ~2 years at Expertia.
  • Led enterprise implementation for 500+ recruiters at Reliance with a successful renewal. Also, highly sticky. Jul 2025: avg WAUs 86.2, MAUs 142. WAUs and MAUs chart for Jul 2025 View zoomed-in visual
  • Shipped AI features across voice calling, resume screening, agentic chat with candidate profile, and interview assistance.
  • Built growth and GTM systems from scratch: analytics, activation, CRM, and ads (100K+ USD topline contribution).

How I AI

  • Building Ek-ai (short for ekagra ai) as a side project - a proactive agent for Indian business operators. Reads WhatsApp, Gmail, and Calendar, with a deep contextual awareness of the business, then emails a morning brief: the 3 things that matter today. Runs unattended on the operator's own Claude or ChatGPT subscription with data staying private. Every run posts a PostHog event - read-through rate and the done-vs-skip ratio on the items in the brief as the feedback loops. Currently piloting with FnF.
  • Built a weekly AI coach that I have been using consistently 12+ weeks in. Started off as an accountability partner of sorts, but now it proposes why I am stuck on some things as a hypothesis, designs small experiments against the blocker, tracks which nudges actually got me to act, and reads my energy before setting next week's load. Three inputs - annual charter, daily journal, weekly plan. Open-source skill.
  • An AI-assisted system of insights from my daily journaling. Implementation backs up Google Docs entries to GitHub, LLM-tags them for mood/ emotions/ people/ themes, emails them back as echoes from 1/ 3/ 6/ 12 months ago, and streams to PostHog for pattern-spotting. Completely free-of-cost to run. 180+ days of journaling streak
  • A job researcher for opportunities I choose to pursue - it does not scout roles. Agents research into the company, role, comp, and interviewers, contrast the opportunity against my interests and background. Everything lands in one composite thread per opportunity: each new conversation, call, or Granola note appends a dated update rather than overwriting, so the doc holds the whole arc via job update agent.
  • Inbox-0, a Claude-native triage app. Claude reads the inbox, decides what needs me, and drafts replies in my voice; a zero-dependency local server renders each one as a card; I click Send via SMTP. Simple workflows surface out follow-ups where a reply is owed, cold outreach, unanswered invite - each with its own silence window and a cap on how many times it may nudge. Cleared cards prune themselves a week later; anything still pending is never touched. Headless CRM system of sorts experiment.
  • I love Granola but their free plan hides notes >30 days old. So I built a system that pulls every meeting - notes, summary, transcript, attendees - and mirrors Granola's own folder structure into a private GitHub repo, auto-syncing daily. I chat with my themed notes from Claude or ChatGPT through the GitHub connector, so past meetings stay queryable instead of expiring.

Current favourite tools

The kind of work I am built for

I do my best work when I can take a real-world problem, make it legible, exercise judgment through experiments, and turn what I learn into working systems.

The roles that fit me best combine:

  • High ownership: I want responsibility for the outcome, not just a narrow slice of execution.
  • Deep work and real-world contact: Time alone to think and build, paired with regular exposure to customers, operators, experts, and live constraints.
  • Ambiguous problems with a concrete test: I enjoy turning fuzzy questions into decisions, experiments, and systems that can be evaluated.
  • End-to-end building: I am skilled at moving between strategy, product, operations, PoC implementation, and the plumbing that makes a system actually work.
  • Small, capable teams: I come alive in environments with low ceremony, direct communication, and room to take initiative and make decisions quickly.
  • Not a production grade engineer for large scale systems.

I am especially drawn to building AI-enabled products and operating systems for businesses: work that sits at the intersection of customer understanding, product judgment, systems design, and execution.

Earlier creative + builder work

Beyond work

Endurance athlete. Learning Hindustani vocals.