NEXA-Sphere: an AI-SaaS concept for multi-outlet F&B inventory, and the road to ICBC 2026 Top 4

The problem

Multi-outlet F&B businesses with 2–10 branches sit in an awkward middle. They are too small for an enterprise ERP, yet too complex for purchasing that runs on instinct. The familiar result is overstock, stockouts, and food waste.

The idea

NEXA-Sphere is a B2B AI-SaaS concept that turns historical point-of-sale data into purchasing recommendations. It is built around three proposed pillars:

• AI demand forecasting (daily and weekly sales per menu item)
• Procurement recommendations with owner approve-or-modify control
• A centralized dashboard across outlets, on mobile and desktop

The owner stays in control: recommendations can be approved or modified, rather than being ordered automatically.

Proposed technical approach

The proposal names time-series forecasting (ARIMA, Prophet) and gradient boosting (XGBoost, LightGBM) as modelling approaches, with cloud deployment on AWS or GCP planned. These are proposals, not a finished system.

What was prototyped

The proposal included interface prototypes built on simulated data: a POS Terminal with an AI Engine analytics view, a Floor Plan Manager and admin dashboard, and navigation tabs for Kitchen Display, Table Map, Data Science HQ, Inventory, Live Chat, and Public Web. A design-partner build, Dimsum Manager with Dimsum Mentai Kemayoran, had role-based login (Owner, Admin, Kasir) running and was about 80% complete at the time of the proposal.

What is not validated yet

• Forecast accuracy (calibration is planned over a client’s first 4–6 weeks)
• Measured food-waste reduction
• Market size, pricing, and revenue projections (indicative only)

Stating this plainly matters. A concept is only as credible as its list of open questions.

On stage at ICBC 2026

The proposal was presented on April 25, 2026 at the International Canvas Business Competition, organized by Universitas Dian Nuswantoro, under the theme “Digital Business Ecosystem for Global Market Expansion.” I presented as team lead for NEXA Tech Labs and the team finished in the Top 4 (Juara Harapan 1). The result is recorded by Kemahasiswaan Universitas Gunadarma on its official announcement. More on the ICBC 2026 page.

My role

Team Lead and AI/ML Engineer: technology architecture direction, predictive-model (demand forecasting) development, and algorithm risk management. See the full NEXA-Sphere project page.