§ 01Snapshot
- Levelclose to ten years in frontend engineering, from junior developer to team lead and technical lead, now a senior hands-on engineer by choice. Works with coding agents every day and builds the checks around them.
- Roles heldsenior frontend engineer at a venture-backed B2B software startup · technical lead at a product engineering consultancy · team lead at a consumer fintech
- Scaleled a small cross-functional product team; mentored and onboarded juniors
- IndustriesB2B software · consumer fintech · client products
- Company typesventure-backed B2B startup · product consultancy · large consumer fintech
- Specialtiestest strategy for apps that have none · end-to-end tests on every pull request · fast CI · data-heavy screens that stay fast as data grows · frontend architecture and refactoring · deterministic guardrails for coding agents · LLM features that can't invent data
- Best forsolo founders with an AI-built React / Next.js app and no tests · seed startups whose dashboards slow down as customers add data · teams adopting coding agents who want checks, not hope
- Core stackReact, TypeScript, Next.js, Redux / RTK Query, TanStack Query, Table and Virtual, Zustand, GraphQL, REST / OpenAPI, Tailwind, shadcn/ui, Playwright, Vitest, Jest, Storybook, Vite, Nx, Turborepo, GitHub Actions, GitLab CI, Docker; Node, Supabase, PostgreSQL; Claude Code (skills, subagents, hooks), MCP, Codex, Cursor
§ 02Services
| ✓ | PR/MR code review | frontend, tests, agent-written code |
| ✓ | Architecture review | frontend |
| ✓ | Architecture design | frontend, test and CI setup |
| — | Audit / due diligence | Not offered by this expert |
| ✓ | Vibe-code rescue | frontend + test coverage |
| ✓ | Team mentoring |
§ 03Track record
A dashboard that took a minute to load, now under a second
Customers with a few thousand items saw the home page take about a minute to load, and the browser used several hundred megabytes of memory. One of our experts rendered only the rows on screen (virtualization) and sent the data requests in parallel instead of one after another. The home page now opens in under a second, with memory use cut by about 80%.
Long histories that froze the browser
Large tables with tens of thousands of rows froze the browser. One of our experts virtualized them and removed the freezes, then cut the app's download size by about a quarter with lazy loading. A size report on every pull request now shows when a change makes the app heavier.
A test strategy nobody asked for, and why it paid off
The product had tests, but no one could say which important flows they protected. On their own initiative, one of our experts listed the critical user scenarios, built a coverage map that updates itself as tests change, and filled the gaps. Critical scenarios covered went from under half to nearly all, and the full end-to-end suite runs on every pull request.
Continuous integration from half an hour to a few minutes
Slow checks made developers wait or skip them. One of our experts cut the unit-test job from about half an hour to a few minutes, halved the end-to-end check and replaced the linter with a much faster one, so every pull request gets the full set of checks without a long wait.
A coding agent with deterministic gates around it
Coding agents wrote useful code but decided on their own when they were done. One of our experts built a ticket-to-pull-request workflow where the agent works in phases, and each phase passes fixed gates before the next: lint rules, hooks, scripts that check the specific change, a model chosen per phase, and a browser check of the ticket's acceptance criteria. They audit which agent instructions actually help by reading the agents' transcripts.
Tickets a coding agent can't misread
Vague tickets produced confident but wrong agent work. One of our experts studied what a well-formed task needs and turned it into a ticket-writing step that investigates the codebase and related repositories first, then writes acceptance criteria in the user's words. A hook checks the criteria are there before work starts.
A complex text-only configuration turned into a visual editor
Customers could only set up the product's most complex feature by editing a text configuration. One of our experts built a visual editor for it, with forms generated from a schema. The visual and text versions stay in sync, so power users and newcomers edit the same thing.
Modernizing a legacy frontend without stopping delivery
An older codebase had loose typing and no shared structure. One of our experts introduced strict TypeScript, enforced code-style rules, moved the code to a feature-based architecture and migrated it to modern tooling in steps, while features kept shipping.
Running a frontend team at a large consumer fintech
A product team needed someone to own planning, quality and people. One of our experts led a cross-functional product team: sprint planning, backlog, technical roadmap, onboarding and mentoring. They set the code-review rules, automated the routine parts of review (commit conventions, automatic pull-request checks) and added performance monitoring with alerts.
Two client apps built from scratch
A consultancy's clients needed new web apps. One of our experts chose the stack, designed a modular frontend architecture and the data-fetching strategy for two applications from scratch, and improved load speed and responsiveness measurably. As technical lead they then set engineering standards and documentation and managed delivery with stakeholders.
An AI report that is only allowed to repeat the math
Personalized reports were built on precise calculations, and a language model invented details when left free. One of our experts made the model write only from the calculation engine's output and added a separate checking agent that verifies every statement against the computed data and a list of banned phrasings before a report is delivered.
Pulling structured records out of uploaded medical documents
People had medical records as scattered photos and PDFs. One of our experts built an LLM pipeline that turns uploaded medical documents into structured, searchable records.
§ 04What you can book
An AI-built app with zero tests: what to cover first
A founder's React / Next.js app was built with an AI coding tool and has no tests. The expert lists the 5–10 journeys that cost money or users if they break, writes end-to-end tests for those first and makes them run on every pull request, before any new features.
A short test plan, the first test suite and a CI check in the founder's repository.
Frontend review of agent-written pull requests
A team ships React / Next.js code written mostly by coding agents. The expert reviews pull requests for what agents get wrong on the frontend: state that goes stale, loading shown as "missing", data fetched twice, lists that will freeze with real data, and download size creeping up.
Review comments plus a short list of recurring patterns to add to the agent's instructions.
Guardrails for a team adopting coding agents
A team has started using coding agents and quality is drifting. The expert sets up the guardrails in the team's own repository: hooks and lint rules that block known bad patterns, checks that run on the agent's change, and acceptance criteria the agent must pass in a browser before it opens a pull request.
A working setup plus a one-page guide for the team.