I'm Hasaam Bhatti, in Toronto. I take products from idea to shelf, from the backend to the box.
Research, ads and sourcing for Amazon sellers. Built because I needed it as one.
Web app
iOS app
Launchie
MCP
Chrome extension
Brand
Sellers use it to find a product worth selling, see what its market earns and check the margin before they order stock, then build the listing and track its rank after launch.
400+ sellers have used it to track more than 1,000,000 markets. I built it end to end: the interface, the API and the data pipeline behind every chart.
Sellers use it to check sales, profit and ad spend for any date range, with a home screen widget for today's numbers.
Launchie, an AI agent in the app, answers questions about the business and proposes ads changes the seller approves on the spot. The app tracks more than $3M in seller revenue. I built it end to end in SwiftUI.
Launch Fast's MCP server brings the whole platform to any AI client that connects over OAuth, like Claude, ChatGPT or Cursor. Sellers research products, check sales and review ads in the chat, and answers come back as interactive Launch Fast interfaces.
200+ connections make more than 100,000 tool calls a month. I built the server, its per-tool OAuth permissions, and the 18 tools with the interfaces they render in the chat.
Launchie reads a seller's own sales and ads data, answers questions about the business and makes changes once the seller approves them. It also runs scheduled briefs, from a weekly profit report to custom checks.
It manages more than $3M in seller revenue. I built it end to end, from the agent runtime on Cloudflare to the approval flow in the iOS app.
I designed Launch Fast down to its type scale, its button sizes and its single blue accent. The same system carries from the product into every email and paid social ad.
The extension adds a search summary to Amazon's results: revenue, price, ad cost and reviews for the market at a glance. The dashboard brings the web app's product, PPC and supplier research onto the same page.
827 sellers use it, and it's rated 5.0 on the Chrome Web Store. I built it end to end on Manifest V3.
My Amazon brand incubator. I find a gap, then design the product and brand that fill it.
Zen Sweat
Flag Runner
A pop-up steam tent with a 1000 W steamer, a chair and a foot massager, set up in five minutes with no tools. I found the gap, then designed the product, its brand, its manual and its box, and had it made and shipped.
A flagpole mount that clamps onto a truck bed with stainless steel screws, so nothing gets drilled, and packs into its own weatherproof case. I designed the mount, its brand, its manual and its box, and had it made and shipped.
An AI studio that makes on-brand posts for CrossFit gyms, built with MediaLaunch.
Chat workspace
Agent interactions
Auto posts
Brand memory
Model choice
An owner describes a post in plain words and gets back one ready to publish, with the gym's own logo, colours and type, so a gym without a designer can still post like it has one.
I built the chat, the agent behind it and the image pipeline.
Every week, Auto Posts makes nine posts on the gym's brand and checks each one before it lands, so the owner opens a finished week instead of a blank page. Their edits teach it what to make next time.
I built the weekly run, the check on every image and the loop that learns from edits.
Most AI design tools start from zero, so a gym owner explains their colours, fonts and tone again with every request. GymCreatives reads the gym's website once and saves its brand, so every post after that looks and sounds like the gym.
I built the brand profile the agent reads before every post, and the rule that it only changes when the owner says so.
Most AI agents either hide what they're doing or bury the owner in text. Here, each step shows up as it happens, results arrive as something to look at, and when a choice is the owner's, the agent stops and asks with a few clear options instead of guessing.
I designed how it feels to work alongside it: every state a step can be in, when it asks instead of deciding, and a one-tap way to hand the choice back.
Image models keep leapfrogging each other, and most tools lock you into one. GymCreatives puts eight models from five companies in one menu, with the credit cost beside each, so a gym can pick whichever suits the post, or whichever is best this month.
I built the registry behind it: each model declares what it can do, and the app adapts its formats, sizes and reference images to match.
A daily hadith on the iPhone, and two menu bar tools on the Mac. All Swift.
Daily Hadith
Beamlet
PR Monitor
My mom listened to hadiths wherever they turned up: forwarded on WhatsApp, buried in YouTube, shared on Facebook. Daily Hadith gathers them in one calm place on her iPhone, with a new hadith each day in Urdu audio, an English translation and a library to come back to.
I designed and built it for her in SwiftUI, from the audio player to the library she browses by theme.
Claude Code's Remote Control lets you keep working from your phone, but a session only starts from a chat already open on your Mac, and it drops when that window closes. Beamlet starts one from the menu bar in a home folder, so Claude can reach every project and open new chats, and it keeps the session running.
I designed and built it in SwiftUI, from the nine-ray mark to how it recovers when the connection drops.
When you code with an agent, it opens a pull request and review bots start checking it on GitHub, but nothing tells you or the agent when they finish. PR Monitor sits in the menu bar, watches checks and review comments across your repos, and tells you the moment a pull request is ready, so the agent can pick up the feedback in one pass.
I designed and built it in SwiftUI, from the state each pull request lands in to when a notification is worth sending.
Smaller tools, agents and automations, built along the way.
Also shipped
2025 to 2026
Smaller tools, agents and automations, built along the way.
