AR & LiDAR Wound Assessment
- Sub-cm
- Distance measurement stability
- 3
- Wound metrics computed
Services · 04
App Maintenance & Code Optimization
Rescue, stabilize and speed up the app you already have, without a risky rewrite.
React NativeReactNode.jsSentry
How this runs
- First deliverable
- A written audit, inside week one
- How work ships
- Small verified steps, ranked by impact
- Downtime
- None — fixes go out as normal releases
- What you keep
- Baselines, dashboards and numbers for stakeholders
Where we start
Three ways projects land on my desk
Most apps do not need a rewrite. They need someone to read the crash logs, measure what is slow, upgrade what is out of support and put tests around the parts that keep breaking. I have done this for apps with hundreds of thousands of users and for small teams who inherited a codebase nobody understood — always the same way: audit first, agree priorities, fix in small verified steps, and leave monitoring behind so the next regression is caught by a dashboard instead of a one-star review.
Your app keeps crashing
The reviews say it, the uninstall curve says it, and nobody on the team can say exactly why. You need the bleeding stopped before anything else.
Week one · Crash clusters triaged from Sentry or Crashlytics, the top crashers root-caused, and the first fixes shipped with tests pinning them down.
Your app got slow
Startup takes forever, screens stutter, the bundle has been growing for years. Users feel it before your dashboards do — if you have dashboards.
Week one · A measured baseline — startup, renders, bundle, API latency — and the quick wins (often config and caching) already in review.
You inherited a codebase
The original developers are gone, the dependencies are years old, and every change feels like defusing something. You need a map before you need fixes.
Week one · An architecture read-through and dependency audit, with the quick wins separated from the structural problems — in writing.
The build sheet
What's on the menu
Codebase & health audit
A written report on crashes, performance, dependencies, security and architecture, with a ranked fix list.
- Report
- Ranked fixes
Crash & bug fixing
Root-cause analysis from Sentry or Crashlytics, fixed with tests so it stays fixed.
- Sentry
- Crashlytics
Performance optimization
Startup time, render performance, bundle size, memory, API latency and caching.
- Startup
- Bundle
Framework & dependency upgrades
React Native, Expo, Next.js, Node.js and library upgrades done in safe increments.
- RN
- Next.js
- Node
Security & compliance fixes
Vulnerable dependencies, auth gaps, data exposure and store policy requirements.
- CVEs
- Store policy
Monitoring & alerting
Crash-free rate, error rate, latency and vitals on one dashboard with sensible alerts.
- Dashboards
- Alerts
Standards I don't negotiate
What an engagement moves
The first two rows are measured ranges from past rescues; the rest are the metrics I baseline in every audit and report on afterwards.
Crash rate
- Before
- Baseline
- After
- Up to 90% lower
App performance
- Before
- Baseline
- After
- 25-40% faster
Production incidents
- Before
- Baseline
- After
- 50% fewer
Cold start time
TODO- Before
- TODO: your baseline
- After
- TODO: measured after fix
Bundle size
TODO- Before
- TODO: your baseline
- After
- TODO: measured after fix
Dependencies out of support
TODO- Before
- TODO: audit count
- After
- 0
Always true, whatever the project
No rewrite required
Incremental fixes inside your existing architecture, so the product keeps shipping.
Measured, not guessed
Every fix starts with a baseline and ends with a number you can show your stakeholders.
Safer with every step
Tests added around each change; upgrades done one version at a time.
Your team levels up
Standards, review checklists and docs that reduced incidents by 50% on past teams.
Proof, from shipped work
Codebases other people were afraid to touch
A clinical imaging app with untyped global state, a statewide field app that had to get 30% faster on 2G, an enterprise app de-duplicated across 15+ modules — rescue work I can show you in full:
Statewide Field Inspection Platform
- 500k+
- Downloads
- 5
- Functionary roles with their own inspection sets
Enterprise Engagement & Wellness App
- 15+
- Feature modules on one UI architecture
- 30%
- Shorter feature delivery cycles
90%
Crash-rate reduction, up to
Measured on rescued apps
25-40%
Performance improvement
Clean architecture, caching, CI/CD
30%
Shorter feature delivery cycles
After de-duplicating 15+ modules
The do-nothing cost is real: Stripe's developer research found engineers lose roughly a third of their week to technical debt and bad code instead of new features.
The stack
Tools picked for the job in front of us
Diagnose
Sentry
Firebase Crashlytics- LILighthouse
Codebases I rescue
React Native- React
Next.js
Node.js
TypeScript
Quality
Jest- DEDetox
Playwright
ESLint
Storybook
Ship
GitHub Actions
Fastlane
Docker
AI in the engagement
Where AI actually shows up
Legacy code is where AI tooling earns its keep: reading large unfamiliar codebases, drafting migrations and generating the tests that were never written.
- 01
Audit
The codebase, mapped fast
Agents map modules, data flow and dead code, so the written audit covers more ground in the same week.
- 02
Stabilize
Tests that pin behaviour
Characterization tests drafted by AI before any refactor, so 'it still works' is verified, not assumed.
- 03
Upgrade
Migrations with a co-pilot
Breaking-change diffs summarized and codemods drafted for framework upgrades, applied one version at a time.
- 04
Triage
Crash clusters, summarized
Logs and crash groups distilled into root-cause hypotheses I verify on device before fixing.
45%
CI time cut by pipelines that follow the diff
Measured on a three-app monorepo
5
Agentic coding tools in daily use
Cursor, Claude Code, Codex, Windsurf, Antigravity
70-80%
Test coverage reached on past codebases
AI-drafted, human-kept
The guardrail · An agent can read ten thousand lines in a minute; deciding which hundred to change is still the job. Nothing lands on a legacy codebase without a test pinning the old behaviour first.
Process
How an engagement runs
Expand each step to see what happens and what you receive.
- Crash and error log review, bundle and startup analysis, Lighthouse and vitals baseline.
- Dependency and runtime audit: end-of-life versions, known vulnerabilities, upgrade blockers.
- Architecture read-through with the quick wins and the structural problems separated.
FAQ
Asked before you had to ask
Almost always maintenance. A rewrite is justified only when the platform itself is dead or the data model is wrong at the core. The audit answers this honestly in the first week, with evidence.
Got an app that crashes, lags or nobody dares to touch?
Send me the repo or the crash dashboard. I will tell you what I see and what I would fix first.