At Pink House Alchemy, Rob worked on moving the retail store from Squarespace to Shopify and adapting the existing wholesale customizations from Odoo 17 to 19.

He customized the commercial Flow theme and built migration tooling for the catalog, including variants, images, and publication state. Preview and single-product modes let him inspect changes before a broader import. His work also covers old-URL redirects, recipe search, and a nutrition-data workflow that previews edits and preserves existing values when a CSV cell is blank.

Wholesale had a rule to carry forward: the combined quantity of eligible products must be a multiple of six, with exception items excluded. Rob adapted the cart behavior, warnings, and checkout block for Odoo 19 and added regression coverage for the rule and the Shopify-to-wholesale handoff.

The retail storefront is live on Shopify. The Odoo work covers the upgrade implementation and regression tests for the existing wholesale rules. Discuss the rules your migration needs to preserve.

Rob built Jobbi for his own job search. The useful distinction is between what a document asks for and what the approved source material actually supports.

Its matching rules select relevant career evidence for a draft. A recognized requirement without supporting evidence becomes a question. The reviewer sees both before using the document, and a saved draft records which claims and source version it used. Agents support the research and preparation around that process. Claim selection and draft assembly are deterministic.

That is a concrete starting point for proposal, intake, or other document work that needs a person to inspect the evidence. Jobbi is a personal product, with human approval retained before submission. Map those decisions to your document workflow.

Rob has built shared UI systems inside existing product teams. He led that work at Brainbase, built the system within his Goldbelly team as an individual contributor, and architected the frontend system at Genius Sports on contract.

His founder work makes the design rules inspectable. Lens, at Clear Labs, gives people and coding agents named, typed choices for components. Unsupported spacing or colors fail type checking, with lint rules checking usage.

Dema, the separate library behind Askrob, uses shared components in the actual application and its reference. Feature-specific layout and animation stay with the feature. A production build once kept the layout but lost some colors, so Rob added checks for rendered color as well as geometry.

An engagement can start with the repeated controls and inconsistencies in your existing product. Work out what to keep, repair, or replace.

Rob built Askrob, his public conversational portfolio, from the chat interface through the model integration. It answers from a small approved knowledge collection, with facts kept separate from voice examples.

The model writes the reply. The application decides which actions are available and when consent is required. Rob built streaming, cancellation, time limits, and failure reporting around generation. Model providers sit behind a shared interface, and an explicit switch leaves an unfinished reply with the provider that started it.

When the system lacks information, the question enters a review process. A visitor cannot turn an unsupported claim into an approved fact by typing it into the chat. These controls make specific behavior testable, while generated answers can still be wrong.

This is Rob's personal public product. Discuss the application your AI feature needs around it.

Rob co-founded Clear Labs and built across the contracts, backend, product interface, and brand. ClearCPU, its publicly launched Solana protocol, is designed for AI agent and compute payments.

That work includes a quote API, a TypeScript SDK, and wallet integrations. When his first price-oracle design left a delay between updates, he changed it to pull fresh prices on demand and built a signing system to verify the data's provenance and integrity.

He also built a browser-wallet testing plugin so connect, sign, and swap flows can run end to end in CI. The interface, the underlying integration, and the checks around it have all been his responsibility.

This is founder-owned product work. Explore the ownership and a possible integration scope.

Rob built the real-time trading interface for MetaPredict, an on-chain prediction market now on mainnet. Prices change while a person decides, signs, and waits for a transaction. The interface has to explain which step they are in.

He separated quoting, signing, swapping, completion, and errors into explicit states. He moved quoting from a server round trip into the client, with user-signed settlement, quote-freshness checks, and configurable slippage. Charts combine initial price history with reactive updates, and browser tests exercise the wallet and swap flows.

That experience applies to products where current data and an action in progress need to stay understandable together. It establishes the implementation, with no measured latency or trading-volume claim. Apply those decisions to your live interface.

Start with one thing a customer or teammate needs to do, what currently gets in the way, and the system Rob would be working in. Share the relevant screen, workflow, or brief and any fixed deadline.

A useful first scope names the behavior to deliver, what it depends on, and how you will review it. That could be repairing a checkout rule, connecting an AI reply to an existing product, or getting shared controls into a working screen. The exact scope and delivery date need to be agreed for your project.

Rob is available to start immediately once scope, access, and required inputs are ready. Book a project conversation.