An application, written as a memo · Chief of Staff to Vinod Khosla

I turn unfamiliar problems into clear decisions.

Staff engineer on payments at RB Global, founder of FarmerCPA, and once before a founder whose company was acquired. Below is the evidence for each thing the role asks for, the framework I use to judge ideas, and three bets I'd defend in front of Vinod.

The 60-second version

  1. Why me 17 years building systems where money moves, from risk platforms to payments, plus two startups, one of them acquired.
  2. How I judge Consequence of success × plausibility, scored on five questions, and I write down what would change my mind before I decide.
  3. Where I'd bet Where AI meets the regulated, physical economy: audit, grid interconnection, and farm labor.

01

The fit, requirement by requirement

Each thing the role asks for, the claim I'm making, and the evidence you can check.

01

Achievement, depth & intellectual range

You've gone deep in at least one area and can learn unfamiliar domains quickly.

I've gone deepest on systems where money and decisions move: risk platforms, same-day payments, ML inference and now settlements.

Seventeen years across e-learning, capital markets, municipal government, banking, retail, aviation, e-commerce ML, marketplace payments and accounting. Along the way: an award-winning risk-management system at IHS Markit, sub-150ms ML inference at Wayfair, and today the settlements architecture at RB Global. I learn new domains formally too: PMP, Google Cloud Architect, and Hinton's and Ng's courses on neural networks and ML.

Google Cloud certification
0industries I've shipped software in

02

Independent thinking

You reason from first principles, distinguish evidence from assumption, and are comfortable saying “I don't know.”

When a process exists only because it always has, I rebuild it from the question it was meant to answer.

At the City of Calgary, cloud evaluations took months. I designed a new assessment strategy around the risks that actually mattered, which cut assessments to three weeks, and drafted the SaaS procurement guidelines and the organization-wide deployment strategy. At IHS Markit, I replaced hand-written code and FpML conversions with generators.

0 wkscloud assessments, down from months

03

Clear communication & people skills

You write precisely, explain complex ideas succinctly, and disagree constructively.

My job right now is getting five teams to agree on one architecture.

At RB Global I run system-design deep dives and present solutions across five payments teams, breaking problems down so teams move faster. At the City of Calgary I brought many cross-functional groups together to adopt new technology. At Wayfair I coached teams on pair programming and TDD, and I mentor new programmers through Chic Geek and HTML500.

This page
0teams aligned on one platform architecture

04

Initiative & disciplined execution

You deliver in an unstructured environment and use AI and automation to improve how work gets done.

I build the automation instead of waiting for it, and I'm doing it now as a founder outside my day job.

FarmerCPA turns bank and credit-card statements into general-ledger entries and one-click financial statements, and AI agents answer accounting questions in plain English. At RB Global I'm introducing agentic coding to the payments teams. Earlier, at IHS Markit, I cut the release cycle from 3–6 months to 2 weeks. This site took a weekend.

FarmerCPA
0 clickfrom statements to financial statements

05

Integrity, composure & low ego

You handle sensitive information responsibly and bring the same care to routine work as to consequential decisions.

Most of my career has been in systems where a mistake costs someone real money.

Payments and settlements at RB Global, same-day money movement for a financial client at Thoughtworks, risk management at IHS Markit, and clients' bank statements at FarmerCPA. Sensitive data is routine for me, and I treat it that way. The low-ego part shows in what I do in my spare time: volunteering on PMI Calgary's Awards Committee and mentoring new programmers.

06

Interest in founders & technology

You want to understand how ambitious ideas become businesses. Venture or startup experience is a plus.

I've founded twice: once to an acquisition, and once still being built.

Parley, a performance-management app, won Startup Weekend Calgary and the Cloud Showdown in Banff, drew multiple investment offers, and was acquired by Uppercut. FarmerCPA is the second, and it has taught me the gap between a working model and a product someone pays for.

FarmerCPA
0companies founded, one acquired

The practical expectations, answered plainly

In person in Menlo Park, 8 a.m. to 8 p.m.
TODO: Yes. I'll relocate to Menlo Park.
Around-the-clock availability
TODO: Yes, and why it suits this chapter of your life.
A three-year commitment
TODO: Yes. What you want to be able to do at the end of the three years.
Visa / relocation
TODO: Your status (e.g. Canadian citizen; TN eligible).

02

Résumé

The short version. Full résumé · PDF ↓

  1. Mar 2025 – Present

    Staff Software Engineer, Payments & Settlements · RB Global

    • Help the payments teams evolve the architecture of the settlements and payments platform, working across five engineering teams.
    • Lead system-design deep dives and break large problems into pieces teams can ship quickly.

  2. Present · alongside full-time work

    Founder · FarmerCPA

    • Built a platform that automates the tedious parts of running an accounting practice. It extracts and maps bank and credit-card statements to the general ledger.
    • Produces financial statements in a single click.

  3. Feb 2021 – 2025

    Lead Software Engineer, ML Platforms · Wayfair

    • Built high-performance serving software for data scientists, delivering sub-150ms latency for large-scale ML inference.
    • Architected an end-to-end testing solution for a real-time fleet-monitoring system that integrates 11 systems, significantly cutting debugging time.

  4. Sep 2018 – Jan 2021

    Senior Software Consultant · Thoughtworks

    • Led DevOps, telemetry and backend development for an event-driven system that enabled same-day money movement for a mid-size financial client (.NET, Azure).
    • Led backend development of an aircraft-maintenance scheduling system for a Fortune 500 client (Java, Kubernetes, AWS).

  5. Oct 2014 – Aug 2018

    Senior Technical Analyst, Cloud Computing & Open Source · The City of Calgary

    • Created the City's cloud-solution assessment strategy, cutting assessments from months to three weeks. Also drafted the SaaS procurement guidelines and the organization-wide deployment strategy.
    • Delivered a learning-management system (Moodle) for 18,000+ employees, saving the City about $2.5M.

  6. Dec 2011 – Sep 2014

    Senior Software Developer · IHS Markit (financial analytics)

    • Led the web team's move to agile, cutting the release cycle from 3–6 months to 2 weeks.
    • Designed a framework for client-specific software packages that helped close deals worth more than $10M.

  7. Jan – May 2013

    Co-founder · Parley (performance management)

    • Won both Startup Weekend Calgary and the Cloud Showdown in Banff, receiving multiple investment offers.
    • Acquired by Calgary-based Uppercut.

  8. Aug 2009 – Nov 2011

    Senior Software Developer · Questionmark (e-learning)

    • Shipped features for the web-based QMLive service in agile, iterative releases that helped the company win new customers quickly.
    • Built multi-language support, including right-to-left languages, opening new markets such as Saudi Arabia.

Education

  • B.Sc. Computing Science · University of Alberta · 2011
  • Computer Systems Technology Diploma · NAIT · 2006

Credentials & coursework

  • Project Management Professional (PMP) · PMI · 2014
  • Professional Cloud Architect · Google Cloud · 2021
  • Self-Driving Car Nanodegree · Udacity · 2018
  • Neural Networks (Geoffrey Hinton) · Coursera · 2017
  • Machine Learning (Andrew Ng) · Coursera · 2016

Side projects

Community

  • Volunteer, PMI Calgary Awards Committee
  • Python and JavaScript mentor, Chic Geek and HTML500
  • Partner, Global Citizens education project, Bali, Indonesia

03

A framework for judging ideas

Most frameworks get used to justify a decision after it's made. This one runs the other way: score each question, plot the idea, and write down what would kill it before falling in love with it. Each question is scored from 1 to 5. Consequence is the vertical axis; the other four average into plausibility.

Step zero · What would change my mind

Before scoring, I write down the one or two observations that would kill the thesis, and how cheaply I could go get them. An idea with no stated kill criteria doesn't get a score.

  1. Vertical axis

    Consequence of success

    If this works, how much does the world change?

    Venture returns follow a power law. A small chance of a very large outcome beats a large chance of a small one, so I size the upside before the odds.

    1 · A better feature5 · A new category, or a scarce resource made abundant

  2. Plausibility

    Kind of risk

    Is the main risk technical (diligence can resolve it) or market (only time can)?

    Technical risk can be investigated: you can talk to the scientists, read the data, and run the test. Market risk ("will people want this?") is cheaper to take after the technology works.

    1 · Unproven demand, commodity tech5 · Hard tech, obvious demand if it works

  3. Plausibility

    Wins on its own economics

    Does it win on cost or performance without subsidies or mandates?

    Policy shifts. Products that are simply better and cheaper survive every administration.

    1 · Needs a mandate to exist5 · Cheaper and better on day one at scale

  4. Plausibility

    Founder–problem fit

    Why is this team unusually likely to learn faster than everyone else?

    The plan will be wrong. The question is whether these founders will notice first and adapt fastest.

    1 · Smart generalists, no edge5 · Earned secrets plus a fast learning loop

  5. Plausibility

    Why now

    What changed in the last 24 months that makes this possible?

    Without an inflection, the idea has probably been tried and failed for reasons that still hold.

    1 · Could have been built in 20155 · Impossible two years ago

Try it

Score an idea of your own, or load one of my three bets.

Ideas plotted by plausibility (horizontal) and consequence of success (vertical) Bet Research Efficient, but small Pass 1 2 3
Plausibility 3.0 Consequence 3 —

← less plausible · more plausible →

04

Three bets

The common thread is AI moving out of chat windows and into the regulated, physical economy: money, power and food.

01AI × Financial services

The audit that never ends

Replace sample-based audits with AI that checks every transaction, continuously.

Consequence of success
4
Kind of risk
4
Wins on its own economics
5
Founder–problem fit
4
Why now
5

What would change my mind: If auditors' liability regimes (PCAOB / state boards) won't accept machine-generated evidence within ~3 years, or if mid-market CFOs won't pay for it before it's required, the wedge collapses.

Read the memo

02AI × Energy

Unclogging the grid queue

AI-native interconnection engineering, which takes power-flow studies from years to weeks.

Consequence of success
5
Kind of risk
4
Wins on its own economics
4
Founder–problem fit
3
Why now
4

What would change my mind: If utilities and ISOs won't accept third-party or AI-assisted studies as inputs (even pre-screening), or if the binding constraint turns out to be physical transformers and transmission rather than studies, this is a consulting business, not a venture one.

Read the memo

03AI × Robotics × Agriculture

Hands for the harvest

General-purpose robotic harvesting for specialty crops, sold as harvest-as-a-service.

Consequence of success
4
Kind of risk
3
Wins on its own economics
3
Founder–problem fit
3
Why now
4

What would change my mind: If pick rate × uptime can't beat a human crew's cost per pound within two seasons of field trials, or if foundation-model manipulation doesn't transfer across crops (one model per crop = services margins), I'd walk away.

Read the memo

Thank you for reading. I'd welcome the chance to be wrong about any of this in person.

abhinav@farmercpa.app