Founder of two shipped products · Bangkok ↔ US-East
AI agents · Automations · LLM integration · MVPs
A working AI build in days — not a quarter-long agency timeline.
I'm Badal Kariwal, founder of Ephoria and SutraHaus. I build AI agents, LLM integrations, automations, and MVPs — fixed price, most builds $2,500–$8,000, delivered in 3–7 days. You don't have to take my word for any of it: every project below is the real, live thing, running inside this page.
Illustrative sessions, real process · first working version in 48h, free
0
years of engineering, from on-device ML to production AI
0
products founded, built & launched end to end
3–7
days from scope call to working software
48h
to a free working demo — you keep it either way
01 — The Proof
Don't read about the work. Click around inside it.
These aren't screenshots — each panel is the live product, embedded. Press interact and use it right here. Spec demos are labeled as spec demos; honesty is part of the service.
№ 1
Ephoria
● Shipped product
Home-care agencies lose caregivers to whoever replies first. Ephoria answers every applicant by SMS in 60 seconds, screens them one question at a time, and books the interview. I founded it and built every layer — product, AI engine, compliance, launch — and the screener you can try on its homepage is the real engine, running in your browser.
EMBEDDED LIVE · try the in-browser screener on its homepage
№ 2
Flowline
◇ Spec demo — built in under an hour
Every data team gets the same Slack message: “is the dashboard up to date?” Flowline is the tool that answers it — freshness against SLA for 22 pipelines, run history, failure drill-downs with logs, an incident view with timelines. Filter to “Failing” and open the settlements pipeline.
The spreadsheet three people edit at once, replaced. Login with roles, full CRUD on an orders table, and an audit log that records who changed what, old value → new value. Sign in as the viewer account and watch every edit control politely refuse.
EMBEDDED LIVE · changes persist locally, one-click reset inside
№ 4
The engine, in the open
● Production code — public
The state machine behind Ephoria's screener, sanitized and published. No freeform chatbot: the model only classifies replies — every word an applicant sees is authored, every guardrail enforced in code. STOP halts permanently. Quiet hours queue. Confusion escalates to a human. The diagram beside this is its actual running order.
My AI creative studio: custom-trained brand models (LoRA), ComfyUI pipelines on RunPod, campaign production for real clients. Proof the design and brand around your build hold up too.
WHY FREE — because a working demo beats any sales pitch, and speed is the one dimension I win on outright. Worst case, you get a free tool.
03 — The Human
The person you talk to is the person who builds.
Founder twice over, engineer for eight years — much of that work under NDA, which is exactly why everything on this page is live and clickable instead.
My proof is what I've founded: Ephoria, a caregiver-hiring SaaS with an AI screening agent at its core, and SutraHaus, an AI creative studio running custom generative pipelines. Shipping my own products means I've done every part of the job — product decisions, architecture, AI, design, deployment, compliance — and it means I build fast, because modern AI tooling in trained hands is a different sport.
There is no team to hand you off to — the person who scopes the build is the person who ships it. I keep US Eastern hours overlap every working day, and I reply fast. If you have a process running on spreadsheets, or an AI feature that never ships: send me two sentences about it.