[work] aquaculture
High-stakes aquaculture ops, made manageable.
MAS Aqua is one of Andhra Pradesh's leading shrimp seed hatcheries: a real technical business run by seasoned operators, over nothing but phone calls and memory. We embedded, extracted the tribal knowledge, and encoded it into a typed operating system with portals, atomic capacity, payment-linked orders, dispatch intelligence, and an agent layer on the same endpoints.
[01] operating reality
Calls, WhatsApp, memory, bank checks, and production Sheets.
Start with respect: MAS is not a company that needed rescuing. It is veterans managing biology, logistics, and money simultaneously, every day. A farmer does not place an order; he calls someone he knows by name. A distributor buys bulk capacity first and names his farmers later. Accounts confirms payments by checking the bank. Dispatch splits by flights and trucks, because the real world is split.
What they were carrying is the tax every traditional manufacturer pays: dynamic data shifting with every invoice, every call, every swing in production volatility. Every step a human. Every handoff a place where the day can go wrong.
[02] the paradigms
Seeds are not SKUs. They are perishable production capacity.
Capacity is booked atomically at the database level: two buyers, one remaining slot, the database decides and exactly one wins. Overselling stops being a discipline problem and becomes an impossibility. Orders confirm themselves when money moves, with advances and partial balances tracked cumulatively across transactions. Dispatch knows the difference between a flight and a truck, because they are different operations with different economics.
The production floor kept its Google Sheets. We made them a live output of the source of truth: workflow untouched, data finally trustworthy. And every action in the system carries who did it, because when livestock and lakhs move daily, the audit trail is how veterans trust a new system.
[03] operator surfaces
Telugu farmers. Gujarat distributors. Local AP distributors. Admin. Flight dispatch. Truck dispatch.
Farmers got Telugu self-service. Gujarat distributors got the two-phase bulk-order and farmer-allocation flow that exists in no software product on earth, because nobody who builds software has sat in that office. Local distributors got the same logic in Telugu with truck delivery. Admin got a single command center: live capacity, payments, reallocations, and audit trails, one screen instead of five phone calls.
[04] agent layer
The dashboard is one client of the 86 endpoints. Agents are another.
The pattern is the same one we run everywhere: the visual surface for operators first, then that same tool wrapped as an agent surface, so agents traverse the exact actions the operators do. A company brain in WhatsApp answers the CEO from live state: what is blocking a delivery date, whose payment is pending, where cash stands.
Autonomous workers handle the routine: collections that knows which buyer always pays late but always pays, capacity protection that nudges pending orders before a hold expires, dispatch resolution that flags tomorrow's missing flight details. And telephone agents answer farmers in Telugu, check capacity, book the order, and confirm on WhatsApp, for the farmers who will never open a browser.
[05] the outcome
219 commits. 86 endpoints. 18 tables. Four portals. One company brain.
Orders that take minutes instead of hours of phone tag. Double-selling that is architecturally impossible. A CEO who asks his phone instead of asking around the office. The operators keep the judgment, the relationships, and the craft. The system carries everything else.
The architecture ports. Poultry, textiles, dairy, agri-inputs, construction materials: every traditional manufacturer running on phone calls and one person's memory is sitting on the same gap MAS was.
[next] explore
Powered by Eagle Eye, Workforce, and The Engine.