2025Operations platform with an AI layer
Restaurant POS + AI Intelligence
A full POS and operations platform — tables, orders, billing, split payments, kitchen display, inventory, branches, riders, reporting — with an AI layer on top for WhatsApp ordering, upselling, demand forecasting, complaint detection and voice ordering.
Status · Full POS and operations platform with an AI layer. Built at CatalystIQ.
The problem
Restaurant groups run operations on a POS that records transactions and nothing else. Ordering, upselling, inventory, complaints and reviews are all handled by people repeating the same judgement calls across branches, inconsistently.
The approach
- 01Build the operational backbone properly: tables and floor, orders, billing, split payments, kitchen display, menu, inventory, branches, riders, employees, reporting.
- 02Add an AI layer on the same data: WhatsApp ordering, menu recommendation, upsell at order time, abandoned-order recovery, complaint detection, review responses, voice ordering.
- 03Forecast demand and inventory per branch so purchasing stops being a guess.
What I did
- Built the POS and operations platform end to end.
- Built the AI layer over it, including WhatsApp ordering and recovery flows.
- Implemented demand forecasting, inventory prediction, segmentation and complaint detection.
Trade-offs I would defend
Operations first, intelligence second
The AI features are only useful because the transactional system underneath them is complete and correct. Building the clever layer onto someone else's partial data was the alternative, and it fails quietly.
Stack & evidence
- Django
- Next.js
- Postgres
- WhatsApp Cloud API
- LLMs
- Role
- Founder / engineer — full stack
- Domain
- F&B operations
Next case study
ATS Resume Generator
LLM-integrated document pipeline