Full-stack engineer who ships in the customer’s environment. I sit with the people doing the work, turn what they actually do into a production change, and debug it live. TypeScript, React, SQL, Stripe, and the APIs already in their stack.
Deployments
Each one started with a customer who was blocked. The write-up is the environment, the integration, and whether they could run it after I left.
Pippa: I embedded with clinicians, then shipped the platform into the clinics. Their feedback was the next release.
MongoTickets: Two customer surfaces, one pool of tickets. I debugged the gap where both of them thought a seat was free.
HerdRight: A marketplace that had to take money and keep prices honest. I put Stripe and a price refresh into the system sellers already used.
Neighborhood Finance: A loan office needed an eligibility answer. I led the deployment and left staff able to change the map without us.
W.R. Berkley: Their underwriters were retyping records. I debugged the live apps and automated the path between the systems they already had.
What I Do on a Deployment
Customer Discovery: Sit with the person who is blocked. Clinicians, underwriters, loan officers. The requirement is the job they repeat, not the first feature they name.
Integrations: REST, GraphQL, and Stripe wired into policy, claims, billing, or a marketplace they already run. No new system of record unless they ask for one.
Production Debugging: Live customer-facing bugs traced across the screen, the API, and the database. The fix ships where the failure actually is.
AI in Their Workflow: An LLM or a vision model in production, with the constraint the customer named. On Pippa the calorie read exists and the patient never sees it.
Adoption: The deployment is done when their staff can run it. Docs, a schema they can edit, and a path that does not require me on the call.
End-to-End Ownership: Ambiguous scope, a data model, the interface, the release, and the number that says it worked. Hours back, dollars saved, or users in production.
Selected Results
Pippa: Shipped a two-sided platform into 200+ outpatient clinics. 10K+ people use it daily. Their feedback became the next production change.
W.R. Berkley: Debugged production across TypeScript, GraphQL, and SQL, and removed a rekeying job that cost 4,500+ hours a year.
Neighborhood Finance: Deployed the eligibility map their staff now edit. The client saves $100,000+ a year and does not need an engineer to change a zone.
HerdRight: Put payments and a four-hour price refresh into a marketplace of 30K+ buyers and sellers.
MongoTickets: Debugged a web app and an Android client against one API so a ticket could not sell twice.
FAQ
The customer is blocked. What do you do?: I go watch the work. At Berkley the request would have been “make data entry faster.” The job was that underwriters were typing the same record into two systems. I automated that path. At Pippa I sat in on 229+ clinician conversations before the backlog was real.
Have you deployed into someone else’s stack?: Yes. At W.R. Berkley the apps had to show the same data as policy, claims, and billing, over REST and GraphQL. At Neighborhood Finance the map had to live where non-technical staff could edit it. I do not stand up a new system of record when the customer already has one.
Have you put an LLM in front of real users?: On Pippa. A vision model reads the meal photo and an OpenAI companion answers the patient. Clinical staff were explicit: the calorie number stays off the patient surface. That constraint shipped in the product, not in a prompt we hoped would hold.
What happens when the requirement is wrong?: I say so, in their words. A form that makes rekeying faster is the wrong requirement. A dashboard the patient can see calories on is the wrong requirement. The written version is the one we build.
What are you looking for?: Forward deployed roles where I embed with customers, ship into their environment, and own the deployment through adoption. Computer Science, Iowa State, May 2026. Based in Chicago.