Rishabh Sharma

Rishabh Sharma

Co-founder of Synthetic Foundry. I lead commercialization and run the business side. I find the numbers that decide the question, then I build whatever moves the answer.

See the workEmail me

I'm building end-to-end software for different industries while taking SynSpect to market at Synthetic Foundry.

Selected work

Open a tile for the full story.

About

I've worked since I was 15.

Rishabh Sharma in a navy suit, arms crossed

I started in food and beverage at Canada's Wonderland and stayed three years. Straight out of high school I went full time at Denny's. Then, going into university, I moved to the buy side at a search fund.

At Boston Beer I managed the finances for one of their Canadian companies and built its cash flow model. I also built a Power BI dashboard that tracked sales and inventory. Production used it to schedule runs and check that materials arrived on time. At Fleet AI I audited and built financial models, and I created and checked more than 600 AI computer-use and tool-use tasks.

Now I lead commercialization at Synthetic Foundry, the weld inspection company I co‑founded. I find the people who need it and take it to market, and I designed the website that sells it.

Most people settle early on either the numbers or the building. I never did.

I like to build software to solve different problems. So far that's a trucking operations suite, a free job search site my friends use and a perks hub for Western's student founders. AI agents I direct write most of it, and checks stop them from shipping broken work.

Outside work. Formula 1, soccer, golf, cooking, EDM, community service and making music in FL Studio.

Experience

Where I've worked. Open any row for the detail.

Building something, or want to talk shop?

Email me about finance, startups, software for small companies or running projects with AI agents.

Email Rishabh LinkedIn
Building now
Building now2026

AxleOS

A small reefer carrier runs dispatch, billing, payables and driver paperwork across spreadsheets, email and phone calls. AxleOS is one system for all of it, built for refrigerated trucking.

Every AI feature makes a draft for a person to approve, and each one has a plain fallback that works without AI. It runs on sample data today. Connections to Samsara and Motive are built, and nothing is wired to a live fleet yet. The driver app and fuel and tax reporting are next. Built with Next.js, React, Postgres and the Claude API.

How it works
  1. Book. An AI reader turns a rate confirmation into a quoted load, and lane pricing suggests a rate from past loads.
  2. Dispatch. A load board moves each load through its stages, refuses to dispatch a truck or driver whose documents have expired, and shows the trucks on a map.
  3. Bill. Invoices, aging, payables and driver settlements, with exports for QuickBooks and Sage.
  4. Ask. A policy assistant answers questions from the company's own documents and cites where the answer came from.
The sales pageSample data
Building now
Building now2026

HotelOS

I'm building hotel management software for independent hotels.

So far I've built the front desk, rate plans and a rate calendar, guest profiles, staff sign-in, night audit, an owner dashboard, and a public booking page that takes a test-mode deposit.

It runs on a sample hotel today, and nothing is live yet. Every screen below shows sample data for a made-up hotel called Northgate Inn.

ScreensSample data
Tooling
ToolingBuilt 2026

Chief of Staff

I run several projects at once. To do that I built a Chief of Staff. It's one chat that knows where every project stands, hands work to AI agents inside each one, and checks what comes back.

An agent can't call a task done until the project's own checks pass, and nothing merges without a review from a model that didn't write the change. Anything the public sees waits for me. This site is built the same way.

How it works
  1. Brief. I describe a task in a sentence. The Chief of Staff writes a brief and starts a session in that project.
  2. Build. Each task gets its own copy of the code and its own branch, so agents never step on each other.
  3. Gate. A stop hook runs the project's tests and page checks. If they fail, the agent goes back to work, up to five tries.
  4. Remember. A wiki updates after every merge, so the next chat starts with what the last one learned. A daily review turns my repeated corrections into rules.
Starter kit on GitHub
Synthetic Foundry Inc.
A scroll through the live home page, with the weld animation playing as you go.
Synthetic Foundry Inc.Designed and built2026

The Synthetic Foundry website

I designed and built our company site, and I wrote it to get quality teams at manufacturers to book a meeting. It has twelve pages, a 3D weld viewer, and a scroll story where a real recorded weld is scanned, measured, judged and recorded as you move down the page. The same story is the site's link-preview video.

I built the site and the animation. I didn't build the inspection software it describes. Our engineering team did. Every number on the site carries a source note, and facts that aren't confirmed yet are marked in the copy, so the build fails if any are left.

How it works
  1. Content. All copy lives in typed files apart from the layout, so the founders can edit words without touching code.
  2. Pages. Next.js and TypeScript render the site. The 3D weld viewer, built with React Three Fiber, loads only when it's needed.
  3. Forms. Meeting, pricing and investor brief requests are validated, then emailed to the team. A booking calendar sits beside the contact form.
  4. Checks. A build step blocks unconfirmed facts, and QA scripts check every page on desktop and phone. It deploys on Vercel.
syntheticfoundry-site.vercel.app Beta

Download it PowerPoint, 3.0 MB
Synthetic Foundry Inc. · Pitch
SFI01 of 12

Matthew Hebbert & Rishabh Sharma

SFI Slide Database

Quality Assurance02 of 12

A Non-Negotiable On The Factory Floor

1 in 3U.S. recalled vehicles never receive the necessary repairs
1 in 5Canadian recalled vehicles never get fixed

Millions of compromised vehicles are on the road today.

Quality must be guaranteed on the factory floor, not patched as an after-sales recall.

Problem03 of 12

Weld Inspection Lags Technological Advancements

CostlyManual Inspections Cost <$200k USD/Year
Inaccurate~10% Of Defects Are Missed In Manual Inspection
SlowIncreasing Cycle Time
Solution04 of 12

SFI’s SynSpect Weld Defect Detection Software

~90%Cost Savings
~99%Accuracy
>1 SecondAnalysis
SynSpect05 of 12

Leverages AI To Provide Advanced Analysis

Scanning
Analysis
Detection
Competitive Analysis06 of 12

SFI's Edge Widens As The Product Line Grows

SFILMICognexCenterlineServo-Robot
Developing weld inspectionYesYesNoNoYes
Deployable softwareYesNoNoNoNo
Willing to partnerYesYesYesYesNo
Machine integrationYesYesYesNoYes
Mobile / handheld deploymentYesNoNoNoNo
Multi-device (phone, edge, PC)YesNoNoNoNo
Multi-standard (CWB, AWS, ISO)YesNoNoNoNo

R&D expands the umbrella: machine integration, mobile & handheld, any device, every code standard.

Bottom rows are on the roadmap: each new capability is another column competitors can't match.

Competitive Analysis07 of 12

SFI Has the Edge Over Current Offerings

The SFI Advantage
  • Synthetic Welding Data Generation
  • Welding Data Access
  • Advanced In-House Model Design
Competitors Challenges
  • Non-General
  • Proprietary Data
  • Off-The-Shelf Models
Product (SynSpect)08 of 12

Alpha User Interface (UI)

The SynSpect alpha interface: a 3D weld view with quality metrics and per-slice analysis.
  1. 1. Weld Quality Metrics. Easy to read and save for compliance
  2. 2. 3D Weld Visualization. Easily analyze and see defects present
  3. 3. Trackable pass/fail statistics. See trends in real-time and act accordingly
  4. 4. Per Slice Analysis. Show exactly where and what is wrong
SynSpect Integration09 of 12

Two Ways To Integrate SynSpect

Option A - API / Edge ServiceOption B - Embedded SDK
How it runsSealed container on the cell PCLibrary linked inside your app
Integration effortLightest, call a local endpointHeavier, in-process integration
LatencyLow (local service call)Lowest (in-process)
UpdatesWe push container updatesYou ship in your release cycle
IP isolationStrongest, model never in your binaryStrong, compiled + node-locked
Best forPartners with their own UI + controllerOne tightly integrated product
Data Security10 of 12

Data Stays Safe With SynSpect

Data Partnership

Anonymized scans flow back to SynSpect

  • The model keeps improving over time
  • You get continuous accuracy gains
  • Governed: anonymized, no end-customer PII, per-site opt-out
Closed Deployment

Nothing leaves the customer site

  • Static model between scheduled updates

Best for security-sensitive customers (defence, automotive)

Expansion11 of 12

One Platform, A Growing Umbrella Of Offerings

Land with weld inspection today. Expand into adjacent AI-inspection products as R&D delivers.

NOWSynSpect core. Structural weld inspection on commodity scanners, the wedge that lands accounts
NEXTMobile & multi-device. Handheld and phone deployment; CWB-certified tier; broader standard coverage
LATERAdjacent AI inspection. New SKUs under the same umbrella, sold through the same channels and relationships
Team12 of 12

SFI Strives To Be A Leader In Computational Weld inspection

Matthew Hebbert

Matthew Hebbert, Founder, CEO

Pursuing: BEng in AI Systems, BEng in Mechanical

Experience: AI design, Automated Manufacturing, Machine Vision, AI Integration

Rishabh Sharma

Rishabh Sharma, CFO

Pursuing: BMos in Finance

Experience: Private Equity, Retail Banking, CPG

Download the deck PowerPoint, 3.0 MB
Equity research · Western Investment Club
Equity researchWestern Investment ClubMarch 11, 2023Buy at C$76.95

Dollarama (TSX: DOL): the call, and what happened

I said buy Canada's largest value retailer at C$76.95. Inflation would push shoppers to value, there was room for hundreds more stores, Dollarcity could take the model into Mexico, and its returns on capital beat every peer. Three and a half years later the stock trades around C$212, up about 175% and well past my bull case.

DOL share price since the buy call
Monthly closeMy case values
Monthly closes from Yahoo Finance (TSX: DOL). The first point is the close before the pitch and the last is November 2, 2026. Case values are from my DCF, with a 2027 exit and an 8% cost of capital. Not investment advice.
What I called, and what happened
  1. Mexico is Dollarcity's next marketSlide 20: zero stores in Mexico, but "a strong candidate for expansion".
    Called itIn June 2024 Dollarama raised its Dollarcity stake to 60.1% and announced Mexico. The first Mexican store opened in 2025, and there were 11 by year end.
  2. Dollarcity is a growth platform in Latin AmericaAbout 350 stores in four countries at the pitch.
    Called itJust over 700 stores by the end of 2025, double the count I pitched.
  3. The cost of living drives salesBase case: about 15% sales growth in fiscal 2024.
    Called itFiscal 2024 sales grew 16.1% to C$5.87B. By fiscal 2026 they reached C$7.26B, from C$5.05B the year of the pitch.
  4. International growth is a catalystSlide 29: Dollarcity "paves the way for further expansion".
    Went furtherIn July 2025 Dollarama bought The Reject Shop, Australia's largest discount retailer, and added 395 stores.
  5. 2,000 Canadian stores by 20301,462 stores at the pitch.
    On pace1,691 stores in January 2026, a little behind my base case of about 1,780, but on track for the target.
  6. Buy at C$76.95, worth C$97.47Bull case C$133.32 at a 22x exit multiple.
    Beat the bull caseAbout C$212 today. My miss was being too cautious. The market paid far more for the growth than the 18x I used in the base case.
  1. Rising cost of livingCPI up 6.8% in 2022, falling household savings and higher rates push shoppers to value retail. 80% of Canadian families live within 10 km of a Dollarama.
  2. Under-penetrated markets1,462 stores against a target of 2,000 by 2030, with Western and Atlantic Canada under-served, plus Dollarcity in Latin America with Mexico next.
  3. Capital efficiencyDirect sourcing and vendor-managed inventory give a 22.7% EBIT margin and 21.1% ROIC, the best in its peer group.
1,462stores at the time of the pitch
~350Dollarcity stores in Latin America
9.3%revenue growth a year over six years
18.0xLTM EV / EBITDA at the pitch
Slides from the deck12 of 40
Capital raising · Emkao Foods
Investor presentationEmkao Foods Inc.April 2024

Emkao Foods: a $XM raise

I built the investor deck for Emkao Foods, which makes Canada's only single-source, traceable organic chocolate in Mission, British Columbia, and trades directly with cocoa farmers.

$13B $5.52B $280M
  • TAM$13 billion. The chocolate market in 2023, using the US as a proxy.
  • SAM$5.52 billion. The share of that market interested in sustainable chocolate.
  • SOM$280 million. What Emkao could realistically capture from its current business.
The ask
$XM

The amount is confidential.

  • Inventory. Ten containers of cocoa beans.
  • Equipment. Replace damaged and faulty machines.
  • People. A production manager, a sales manager and other staff.
  • Marketing. Campaigns to grow sales and recognition.
Slides from the deck7 of 11
Free for friends
Free for friendsBuilt 2026

Job Pipeline

I built it for myself first. Then my friends wanted it, and most of them can't set up code, so I turned it into a free website. A friend signs in with Google, adds a resume, picks roles and cities, and gets their own dashboard of jobs scored against that resume.

It covers 50 role types, from finance and startups to law, nursing, teaching and the trades, and anyone can add a role that isn't listed. Each person's list, statuses and notes are theirs alone. 368 tests run before anything ships.

How it works
  1. Sign in. Google sign-in, then a short setup: resume, roles, cities, and whether you want internships or full-time.
  2. Search. Jobs come straight from company job boards, with their posted dates. Duplicates are dropped.
  3. Score. Each job is scored against your resume, with reasons, and filtered by level, start date and industry.
  4. Apply. Resume Studio reviews and tailors your resume for a role, and the tracker keeps statuses and notes per job.
job-pipeline-site.vercel.app
Game
The promo, cut from the real TV and phone screens.
GameBuilt 2026

The Haunted Manor

A multiplayer party game. A TV runs the show and each team plays on a phone. Teams move through fifteen rooms answering questions, and wrong answers cost hearts.

How it works
  1. Host. A Node.js server runs the game. The TV shows a room code and a QR code.
  2. Join. Each team scans the code and plays on a phone.
  3. Play. Every room gives 30 seconds for a question. The scares fire on every screen at the same moment.
Consulting
Consulting2026

EZ-AI Integrations

AI and online sales setup for small businesses in London, Ontario. We knock on doors, find what costs the owner the most time, and build the fix. That's bookings, review replies and the admin that eats their evenings, and for local retail shops, online ordering, delivery and shared books.

For retail I built a platform where customers browse live stock and order, owners see sales and margins in real time, drivers get a portal and each partner's share of the stock and the cash is tracked. It uses Gemini in Google AI Studio, React, TypeScript and Firebase. The clients, their products and their numbers stay private.

How it works
  1. Walk in. We visit shops in person and ask the owner where the week goes.
  2. Find the cost. A short report names the tasks that cost the most hours or lost customers.
  3. Set it up. Booking, review replies, after-hours calls, quotes or an online store, built to fit how the shop already works.
  4. Show the return. A monthly report counts the hours saved against a baseline taken before the setup.