Fassix · an AI experiments lab
Fassix is an AI experiments lab. We pick an industry, build AI software for it, and run it with real users to see what changes.
What we build
Not a list of skills. These are things we run in production today, each one built end to end.
01
A WhatsApp platform that has handled peaks of 1,700 customer messages a day for a single franchise client, with the AI replying inside two seconds at the 95th percentile.
02
Businesses connect their own WhatsApp and Instagram numbers through Meta’s embedded signup, on infrastructure Fassix operates as a registered technology provider.
03
Every model call is metered, priced and settled against the customer it served. An agent that cannot account for its own cost is a demo, not a product.
04
Razorpay checkout, signed webhooks, idempotent settlement, monthly invoices and GST-correct bills of supply, generated and reconciled without a human in the loop.
05
Career reports rendered as print-quality PDFs in English, Hindi and Gujarati, with the shaping and fonts Indic scripts actually need, not a font swap.
06
A Flutter app for the team on the road, with notifications delivered straight through Apple and Google’s transports from our own servers. No third-party push vendor in the path.
Where this goes
Not chat. Agents that answer a customer, qualify a lead, counsel a student, and can be held to what they said. That is the layer we are here to get right.
Reliable, observable, boring where it should be boring. Systems a small company can run for years, and a large one would recognise as serious.
We build and run our own products. We do not take on client work. If you want what a product does, it is yours to use. If you want it built for you, that is not what this is.
About
Engineers. Based in India.
We are engineers. We like problems that only give way to careful systems: messaging that must not drop a customer, money that must reconcile to the rupee, models that must be right often enough to be trusted and cheap enough to be sold.
Fassix is where we put that to work. Every experiment is built end to end, from the database to the screen a customer touches, and run in production with real businesses and real students on it. When something breaks at two in the morning it is our phone that rings, and that is how we want it. The products get better because the people who built them are the people who answer for them.
Marketing and education came first because both run on people’s time and attention, and both have been badly served by software so far. There will be more experiments.