Building reliable, business critical software for 9+ years.

JotGo is a SaaS platform and browser extension for bug reporting and product feedback, with screenshot capture, session replay, and team collaboration. I lead the rebuild as the main engineer and own the whole system: the database, the backend, the frontend, the browser extension, and the release process, along with the direction of a small three person team. I inherited a rough codebase, so a large part of the job has been restructuring it into something maintainable while still shipping new features.

• Rebuilt the feedback canvas from ReactFlow to Pixi.js and WebGL. It handles hundreds of images smoothly in production, and I stress tested it past a thousand to confirm there was headroom.
• Built the browser extension diagnostics in the background service worker: network request capture, console log capture, and session replay, so developers get real context to reproduce an issue.
• Built the backend on NestJS and Firestore, with GraphQL APIs, OAuth authentication, and Redis and Bull queues for background jobs and email notifications through SendGrid, all running on GCP.
• Integrated Linear end to end over GraphQL and extended an inherited Jira integration, connecting user feedback to the teams that act on it.
• Built an MCP integration that connects the platform's feedback data with AI tooling, supporting both OAuth and PAT authentication.
• Own the git branching strategy and the release process for the team.

Cadent's platform is an enterprise DSP for programmatic advertising, covering campaigns, line items, ad groups, targeting, optimization, and audiences. It's a large, data heavy application, the kind with big server paginated tables and deeply nested, interdependent forms. I worked mostly on the ad group side, and I spent most of my time in the optimization area, which is the most complex part of the product. This is also where my professional React experience started. When a trader sets up an ad group, they pick an optimizer, predictive models, KPIs, and a bidding strategy, and all of those choices constrain each other, so changing one value changes what is valid everywhere else. I was the primary developer on that flow. I wrote most of the optimization form, around 78 percent of a component that had grown past 1,600 lines, and I owned the shared state that ties its steps together. Much of the real work was untangling edit mode and replacing fragile patterns so invalid combinations cannot be saved.

• Owned the ad group optimization flow (optimizer, predictive models, KPIs, bidding), built with React Hook Form and Zod for the interdependent validation and TanStack Query for the many dependent lookups.
• Built the predictive KPI card and the HCP and health audience selection and save flow end to end, using Material React Table for the data grids.
• Built the Next.js API routes, a BFF layer between the React app and the .NET backend, that handle auth, route to the right service, and reshape the data.
• Managed shared state across the app with Zustand, and shipped changes behind DevCycle feature flags for gradual, controlled rollout.

AdTheorent's DSP was an earlier AdTech campaign management platform, with campaign setup, line items, targeting, reporting, optimization, advertisers, and audiences. I worked full stack here for about three and a half years. On the backend I built Web API endpoints and controller actions in C# and .NET, and wrote a lot of SQL and stored procedures with ADO.NET for the data heavy workflows. On the frontend I worked in JavaScript and jQuery. I also spent real time refactoring fragile areas and tracking down production defects across the API, auth, and data. This is the backend heavy foundation of my experience.