Skip to main content
Williams BalogunGet in touch
Case Study — Light / 03

Dexla

An AI-powered app builder: drag, drop and describe your way to a real product, with the editor doing the heavy lifting.

Role Software Engineer · joined as JuniorTimeline 2023 — 2025Type AI builder · WebStatus Archived
dexla.ai — editor
Dexla, an AI-powered drag-and-drop app builder
01 — At a glance

Dexla let technical and non-technical founders ship real apps fast: a drag-and-drop editor with AI woven through it, from generating layouts to fixing bugs. I joined as a junior engineer alongside three seniors, starting on bug fixes and new canvas components, then growing into owning the builder's performance.

The product stayed in closed beta: a few dozen testers from the waitlist and three paying customers. I built the MVP for all three on the tool itself, working directly with each founder. One of them, Airstride, later hired me.

Role
Software Engineer (joined as Junior)
Duration
2023 — 2025
Team
3 senior engineers + me · 2 designers · 2 PMs
Stack
Next.jsTypeScriptMantineReact QueryReactflowPythonFastAPI
3
paying customers, whose MVPs I built on the tool
25%
faster initial load (3.2s → 2.4s)
30%
fewer redundant network calls
02 — What I built

The hard parts.

01Custom drag-and-drop engine

Contributed to a bespoke drag-and-drop library for the editor, adding components and fixing the sync bugs that kept real-time state consistent while several people edited at once.

Drag & dropReal-time syncReactflow
02AI woven into the canvas

Integrated Claude & OpenAI for context-aware actions (change design, fix bugs, connect data) driving a 35% lift in engagement after release.

Claude APIOpenAI APIContext-aware
03Pixel-perfect from hand-off

Translated complex Figma hand-offs into a fluid Mantine UI, the detail that earned repeated praise from designers and PMs for fidelity and polish.

Next.jsMantinePixel-perfect
04Performance & delivery

Cut initial load 3.2s → 2.4s (25%) with code-splitting and render optimisation, and removed 30% of redundant network calls by unifying endpoints behind React Query caching.

Code splittingReact QueryCaching