Skip to content

Real software built by Ileen

Real builds, not demos.

Ileen has generated production-usable software from technical briefs, worked across existing codebases, and deployed applications to cloud environments. Below is the full library of public case studies. Names are anonymised, the work is real.

Autonomous build. Human-controlled release.


Headline cases

These three are the proof points to start with. They cover the full delivery chain: design the architecture, build the code, deploy to production.

01 — Discovery → Production in 11 hours

Image-to-3D + BOM pipeline →

A real technical discovery turned into a complete AI pipeline: frontend, backend, database DDL and a production-usable MVP. The architecture was revised mid-conversation from fragile semantic search to a DB-grounded agentic flow.

~10h autonomous · ~1h human deploy · frontend + backend repos · DDL generated


02 — Codebase → Live GCP in <10 minutes

Deploy Mode: Sports platform on Cloud Run →

Existing Next.js + Supabase + Stripe + OpenAI app. Ileen analysed the codebase, detected 13 missing secrets, provisioned them in Secret Manager, deployed to Cloud Run and validated HTTP 200 — then guided a non-technical user through every third-party account setup.

<10 min active deploy · 13 secrets provisioned · HTTP 200 validated


03 — Multi-repo CRM with AI agent layer

Develop Mode: Agentic CRM →

Real production codebase. Three repositories. 11 tasks executed across ~3 weeks — feature builds, real bug fixes (soft-delete reads, sync issues), security patches, and a greenfield admin frontend. Develop Mode entered an unfamiliar codebase and made targeted changes without breaking what worked.

3 repos · 11 tasks · bug fixes + security patches · greenfield frontend


All case studies

Design Mode

Design + Build

Develop Mode

  • Agentic CRM → — Multi-repo CRM with AI agent layer, 11 tasks across 3 weeks.

Deploy Mode

Content Mode


How to read these case studies

Every case study follows the same structure so you can compare them like-for-like:

  1. At a glance — facts and numbers up front.
  2. The brief / project — what the user actually asked for.
  3. What Ileen produced autonomously — concrete artefacts, no abstractions.
  4. What required human action — explicit, not hidden.
  5. Honest limitations — what the system cannot do today.

This is not marketing. It is documentation of real work. If you find a claim that looks impressive but is not traceable to an artefact, please flag it.