## Problem ForkNKnife is a restaurant POS used across dozens of locations: order entry, kitchen routing, payments, and customer-facing ordering via WhatsApp. The team needed throughput (10,000+ transactions a month at peak), live visibility for owners, and a way to take orders without making customers download yet another app. ## Approach Node.js services behind a REST API, MySQL for the operational data, WebSockets for live order tracking on the kitchen and dispatch screens. WhatsApp integration via Dialogflow NLP so customers could type natural-language orders and the system would parse, confirm, and route them. Analytics dashboard built on the same data so owners could see daily sales without a separate BI tool. ## Outcome 10,000+ monthly transactions processed reliably; live order tracking adopted by the kitchens; the WhatsApp channel became a meaningful share of incoming orders within a quarter. ## Role Backend and integration engineer: API design, real-time pipeline, WhatsApp/Dialogflow wiring, dashboard backend. Worked with a frontend team that owned the POS UI and a customer-success team that onboarded restaurants. ## What I learned NLP for "place my usual order" is much easier when you give Dialogflow a tight intent schema and let the app handle disambiguation; trying to make the model do everything fails noisily. WebSockets are great until your reverse proxy times out the long connections — health checks and reconnect logic earn their keep.

Type: Enterprise · Year: 2022