Pour temp 1,084.6 °C — Cu · Independent software foundry
The independent foundry for AI-built software. We pour new AI-native products, take prompt-built prototypes to production, and keep both alive — audited, hardened, and verified by a human who signs their name.
Two doors, one production line
Door 01 — The Cast
An agent assembly line builds it: a PM agent drafts the requirements, an architect agent designs the data model, engineer agents write the code, QA agents try to break it. Every stage ends at a human gate — nothing ships unreviewed. Fixed-scope MVP sprints. Born forged: verified against the Zha Standard from day one.
PRD ⌖ → SCHEMA ⌖ → CODE ⌖ → TESTS ⌖ → DEPLOY ⌖
Door 02 — The Forge
Your app works in the demo and breaks everywhere else. We refactor the architecture, harden auth and secrets, build the test suite, containerize it, wire zero-downtime deploys, and cut the token bill. Specialty: Next.js + Supabase apps from Lovable, Bolt, v0, and Cursor — the stack most AI tools generate.
INTAKE → AUDIT → HARDEN → VERIFY ⌖ → SHIP
The reality check
| Raw material — the vibe-coded input | Cast metal — the foundry output |
|---|---|
| "Works on my machine" local environment | Cloud-native, auto-scaling architecture |
| Brittle code and silent runtime failures | Engineered test coverage, verified before handover |
| Exposed API keys and prompt-injection risk | Hardened security, encryption, and proper auth |
| Token-bloated, expensive LLM calls | Optimized model routing and semantic caching |
The Forge — productized, never hourly
Know where you stand
Fixed quote with your free assay
Make it safe to grow
Fixed quote with your free assay
Make it production
Fixed quote with your free assay
Foundry Care
Software rots from day one — dependencies decay, CVEs appear, costs drift. Care is a monthly subscription: our agents watch, patch, and report; a human signs every month's health check.
The production line
Stage 01 — Intake & Audit
A repo link or an idea brief. We assay what exists and scope what should.
Stage 02 — Blueprint
Requirements broken down, production schema designed, scope fixed in writing.
Stage 03 — Assembly
Engineer and QA agents write, test, refactor, and re-test until every assertion holds.
Stage 04 — Human Review
Every line the agents produced gets read, challenged, and approved by a human engineer.
Stage 05 — Shipment
A running production environment, handover docs, and the option to keep us watching.
Positions
Every engagement is a fixed scope with a fixed quote, agreed before work starts. The meter never runs.
Agents build fast; humans decide what leaves. Every stage of the line ends at a human gate.
The Zha Standard is public. Check anything we ship against it — that’s what it’s for.
No retainers that drift, no phase twos that never end. A scope closes, the metal ships.
The Zha Standard
Every project leaves the foundry checked against this list — and you can check it too. That's the point of publishing it.
Zha Standard v0.1 — expands as the foundry learns
Teardowns
Before, after, and what it took — real repos, real failures, real fixes. The first castings are on the floor now.
Teardown Nº 001
In the crucible
Teardown Nº 002
In the crucible
Teardown Nº 003
In the crucible
The founder
Prem Kannan
Founder · Coimbatore, India
I'm an integrated MSc Data Science student who builds AI products in Python. At Bloom Matrix I co-built Anril AI, and I write the evaluations and tests that catch regressions before users do. The Zha Foundry applies that discipline to software that was built with AI and now has to survive real use.
thezhafoundry@gmail.comFinal pour
Let's turn your vibe into structural steel. Send a repo, a broken deploy, or three sentences about an idea — a human reads every message and replies within one working day.