SENIOR PRODUCT MANAGER · PSPO CERTIFIED · TORONTO

IDEA TO LAUNCH

Product Development · Product Delivery · Product Management
backed by Data · powered by AI

Malay Sherasia is a product leader based in Toronto, Canada, with fourteen years across marketplaces, financial services and early-stage ventures.

I have run consumer web and mobile products, migrated enterprise websites for large banks, and founded a funded consumer start-up that grew to 200 cities.

See the work

0

lead volume in 90 days after I shipped save on the search results page

0

pages moved in an enterprise website migration, on a site with 10M+ monthly visits

0

less time developers spend hunting through Jira, thanks to an AI agent I built

0

drop in fuel theft for an ERP client, worth about $95K a year

0

cities in India, reached by a start-up I founded and later exited

FIGURES COME FROM SHIPPED WORK. WHERE I DON'T HAVE A DEFENSIBLE NUMBER, I LEAVE IT OUT.

TWO JOBS I DO WELL

Some teams need someone to decide what gets built. Others need someone to get a committed plan over the line. I have done both jobs, often at the same time.

Senior Product Manager

Decide what is worth building

Turn customer behaviour and funnel data into a short, ranked roadmap that engineering can commit to.
Write the problem down before the solution. PRDs, acceptance criteria and success metrics agreed up front.
Put AI where it removes real work, with guardrails for what happens when the model is wrong.
Work across web, iOS, Android and lifecycle email as one product.

Senior Product Delivery

Get it shipped on the date

Run refinement across hundreds of epics with dev teams, agencies, UX, QA and the business at one table.
Freeze scope at a known point so the last stretch goes to regression, sign-off and cutover rehearsal.
Manage external vendors with specs and outcomes, and leave the craft to them.
Keep compliance gates (AODA, OMVIC, PIPEDA) in the plan from day one instead of at the end.

THINGS I HAVE SHIPPED

Six products I founded, led, contributed to or wrote. Each one is live, so you can open it and check.

PROJECT · APPLIED AI 9ine9ine.com ↗

9ine9ine

A call-intelligence product for local service businesses. It turns phone conversations into clear summaries, so a team knows what each caller needed and what to do next.

THE APPROACH

Product: Putting AI to work on a real business workflow means deciding where it helps, where it needs checks, and what a busy owner actually wants to read.

Delivery: Working with design and engineering partners and modern AI tooling. Clear specs, small releases and a plan for rolling back.

PROJECT · CONTRIBUTOR · AUTO RETAIL cdd6.ca ↗

cdd6.ca

A deals microsite for a Toronto dealership. I worked on the build and search visibility using an AI workflow, and the site was ranking in search within 90 days.

THE APPROACH

Product: Phone calls were the conversion that mattered, so the site was kept call-first and success was judged on real calls instead of click counts.

Delivery: I worked inside a regulated setting: deal pages follow OMVIC rules and get dealer sign-off before release, and one data file keeps deal changes consistent across pages.

OPEN-SOURCE FRAMEWORK Read it ↗

The Final Third

A rule for delivering when the date will not move. It reads Jira or a spreadsheet export and works out the exact freeze date for a release.

THE APPROACH

Product: I can turn a messy argument about scope into a rule that stakeholders, auditors and engineers can all explain. I chose plain rules over a scoring model because a score would have looked more precise than it was.

Delivery: The last third is reserved for burn-down, full regression, defect fixes, sign-offs and a rehearsed cutover. Nothing new enters.

OPEN-SOURCE TOOL claude-repo.com ↗

never-again

Turns a repeated coding mistake into a guardrail that AI coding agents cannot skip. If a script can catch it, it becomes a hook. If not, it becomes one line in a capped lessons file.

THE APPROACH

Product: I build for how AI agents fail in practice. Warn before block keeps trust, and fire logs show which rules earn their place.

Delivery: A retro action that enforces itself. Lessons stop living in a document nobody reads and start running at every commit.

FOUNDER · FUNDED START-UP · EXITED

midnightcake.com

A cake delivery start-up I founded and grew to 200 cities across India. Every cake was baked to order and delivered within three hours.

THE APPROACH

Product: I found a promise customers would pay for, took it from one city to 200, and earned start-up funding along the way.

Delivery: Keeping a three-hour promise while adding city after city is a delivery problem that has to be solved again every day.

FOUNDING TEAM · B2B SAAS ERP orecs.com ↗

ORECS

A cloud ERP for construction machinery rental and contracting businesses. I led it end to end with a team of seven, from client interviews and architecture through on-site implementation and launch.

THE APPROACH

Product: I found a costly problem during client interviews and shipped a module that paid for itself, then sold it to new customers.

Delivery: I carried a multi-module product through requirements, design, development, testing and on-site go-live with a seven-person team.

WHERE I HAVE DONE IT

Fourteen years of product and delivery work, from a founding team in India to large Canadian organisations.

Product manager, web and mobile

Online marketplace, Canada

Merged two competing save actions into one, then shipped save on the search results page. Lead volume rose 14% in 90 days.
Owned the consumer site and the iOS and Android apps, including the launch of a personal garage feature.
Led a lifecycle email platform migration, covering IP warm-up, domain reputation and DMARC.
Built an AI agent over every Jira epic and story. Developers get answers in minutes, spending about 60% less time searching.

Product manager, enterprise website

Large Canadian financial institution

Delivered a milestone migration of 3,000+ content-heavy pages to a new enterprise CMS, on a site with 10M+ monthly visits.
Launched a financial wellness content platform.
Ran weekly refinement across hundreds of epics on Kanban with developers, agencies, UX, QA and business owners.
Won an internal hackathon with interactive analytics dashboards.

Digital product journey manager

Large Canadian financial institution

Launched a banking microsite for newcomers to Canada.
Used journey analytics and accessibility testing to decide what to fix first, with AODA built into the work.

Digital product lead

Automotive retail group, Greater Toronto Area

Led the end-to-end build of an online car financing application, the group's first.
Grew the website into the source of about 75% of sales. Organic traffic rose 26% and leads 13%.
Managed about $50K a month in Google and Meta spend.

Product lead, founding team

ORECS, B2B SaaS, India

Owned client interviews, requirements, architecture, design, testing and on-site rollout for a multi-module ERP.
Led a team of seven developers and designers.

Founder

midnightcake.com, direct-to-consumer, India

Founded an online cake delivery start-up with a three-hour delivery promise.
Scaled it to 200 cities across India.
Selected and funded as a start-up.
Exited the business and kept the brand and domain.

HOW I WORK

These are the habits behind the results above. Each one came from a project that went sideways first.

THE STARTING LINE

Identify the problem, backed by data

Before jumping to a solution, I find the actual pain point using funnel metrics and customer behaviour. If the data doesn't support the problem, we don't build it.

FROM THE FINAL THIRD

Freeze at two thirds

Every late yes pushes the deadline. After the freeze, a new idea waits for the next release.

FROM AUTO RETAIL

Measure what pays

Click events can be inflated by bots. I check them against real calls and sales before calling anything a win.

FROM THE FINAL THIRD

Rules over scores

If a client, auditor or regulator has to understand the answer, a clear rule beats a model with false precision.

FROM VENDOR WORK

Specify the outcome

Vendors and agents get the problem, the constraints and the acceptance criteria. How it looks is their craft.

FROM NEVER-AGAIN

Make lessons enforce themselves

A retro note is easy to ignore. A check that runs on every commit is not.

EVERYWHERE

Every claim survives a check

If I can't defend a number in a room, it stays off the slide and off this page.

TOOLKIT

The platforms and tools I use to plan, build, measure and ship.

Product and delivery

Jira and Confluence
Scrum and Kanban
Roadmaps, PRDs and acceptance criteria
Release and cutover planning
Vendor and agency management

AI and agents

Claude Code and the Claude API
ChatGPT Enterprise and Codex
OpenAI API
RAG and vector search
Model Context Protocol (MCP)

Data and insight

GA4 and Adobe Analytics
Power BI and Tableau
Looker Studio and Amazon QuickSight
Funnel and journey analysis
AI-assisted analysis in Python

Cloud and web

AWS and Cloudflare Workers
Next.js, Vercel and Supabase
GitHub
Enterprise CMS platforms
WCAG and AODA accessibility

Growth and CRM

Google Ads and Meta Ads
Salesforce Marketing Cloud
Lifecycle email and deliverability
SEO and structured data
Conversion tracking