Make the whole team faster.
Led three frontend teams while staying hands-on. Better delivery infrastructure, more reliable forms and shared engineering practices that multiplied across the organisation.
Frontend quality depends on more than individual screens. Deployment time, form reliability, flow logic and inconsistent tooling all affect how a team ships. My work at Sense combined hands-on engineering with practices that could spread across teams.
I managed three frontend teams of three engineers each and contributed as an individual engineer on one of them. I owned code-review standards, technical solutioning and company-wide code-quality decisions, alongside deployment and frontend tooling work.
A map of my contribution areas, not a reproduction of Sense’s deployment pipeline.
S3-based deployment architecture redesigned; reported deploy time improved from 20 minutes to under 60 seconds.
React Hook Form and Zod introduced with hands-on adoption sessions and pairing; reported form bugs reduced by 80%.
XState introduced to make flow-based application logic explicit through state-machine design.
Vite, Biome, review standards and technical solutioning helped spread common practices across the organisation.
I redesigned the S3-based deployment architecture. Reported deployment time fell from around 20 minutes to under 60 seconds, alongside a 70% AWS cost reduction and 30–80% performance improvement.
Deployment speed and infrastructure cost were addressed together at the architecture level. The internal topology and measurement data are not reproduced in this public case study.
I introduced React Hook Form and Zod organisation-wide, with adoption sessions and pairing involving nearly every engineer.
The work included both implementation and helping engineers use the new approach. Form-related bugs fell by a reported 80%.
I introduced XState for flow-based application logic and modernized the toolchain to Vite and Biome. I also contributed to architecting an organisation-wide AI code-fixing tool.
State-machine-based flow logic, shared review standards and common tooling gave teams more consistent engineering foundations. The AI tool was a contribution, not a sole-authorship project.
All numerical outcomes are candidate-reported from my approved résumés. The before/after comparison describes results; it does not imply a specific unpublished infrastructure topology. No employer code, internal screenshot or customer data is included.
From an idea to a published story.