AromaSys is an enterprise-grade warehouse intelligence platform built for Sima Arome, a leading Indonesian natural extracts manufacturer powering F&B, cosmetics, and wellness brands. Watch our demo to see how we bridged fragmented, traditional workflows into a unified, high-efficiency digital twin ecosystem.
🔴 Problem Statement
Operating a premium natural extracts facility requires extreme operational precision, but Sima Arome’s daily workflows were bogged down by manual, fragmented processes. Operators found themselves wasting hours on double-entry tasks—copying incoming batch logs across paper notebooks, Excel spreadsheets, and independent legacy apps. This disjointed flow resulted in frequent data transcription errors, lost tracking details, and severe operational opacity where schedules and batch records lived only inside people's personal chats and pocket notebooks.
Furthermore, Sima Arome faced serious quality control and storage vulnerabilities. Botanical raw material grading relied completely on a few expert workers, meaning that throughput stalled immediately when staff were absent. In the cold-chain warehouse, critical temperature monitoring (-4°C to -20°C) was logged manually on spreadsheets, leading to late anomaly detection and costly batch spoilage. Sima Arome needed a solution that would eliminate manual bottlenecks, secure their cold chain, and bring total visibility to their manufacturing floor.
🟢 What it does
AromaSys consolidates the entire warehouse and production ecosystem into a single, intelligent operations console. Instead of disjointed processes, the platform provides a unified workspace with six core capabilities:
- 📊 Centralized Analytical Dashboard: Surfaces active stock counts, capacity utilization, expiring lot countdowns, and cold-chain warnings at a glance using interactive charts.
- 🗺️ Interactive Floor Plan (Digital Twin): A drag-and-resize warehouse floor map that reads uploaded blueprints via Gemini Vision, automatically parsing layout zones into visual, coordinate-mapped interactive displays.
- 👁️ Dual-Model AI Quality Control: Uses computer vision models to scan incoming raw plant and fruit materials, automatically highlighting surface defects and bounding boxes with an integrated manual inspection failover.
- ❄️ Real-Time Cold-Chain Telemetry: Continually tracks zone temperatures, rendering svg sparkline trends, and automatically generating maintenance tickets if anomalies exceed thresholds.
- 💬 Live DB Context AI Copilot: A Gemini-powered operations chatbot that receives real-time database snapshots to answer complex inventory queries, suggest optimal storage slots, and generate PDF reports.
- 📁 Smart Data Ingestion: Simplifies material intake by using AI-driven OCR to parse unstructured PDFs, CSVs, and documents, automatically detecting duplicates before database entry.
🛠️ How we built it
We wanted a system that was robust, lightning-fast, and secure. We structured AromaSys as a high-performance monorepo, maintaining a strict separation between our frontend client and backend API server to ensure clean system boundaries and simple scaling.
- Frontend: Next.js (App Router) and React written in TypeScript, using Tailwind CSS for premium glassmorphism styling, Framer Motion for micro-animations, and Recharts for data analytics.
- Backend: Express.js running on Node.js (ESM), featuring full server-side JWT authentication, secure Role-Based Access Control (RBAC) middleware, and express-validator input sanitization.
- Database: PostgreSQL hosted on Neon Serverless, leveraging optimized, parameterized SQL queries and connection pool management to ensure high performance and prevent injection attacks.
- AI Integrations: Google Gemini 2.5 Flash API (for PDF/blueprint OCR and contextual chatbot) and Roboflow Inference API (for real-time computer vision grading).
⚠️ Challenges we ran into
Integrating live relational database states with conversational AI models proved to be our biggest hurdle. Injecting an entire database schema and active inventory tables into Gemini's prompt context without blowing past rate limits or inflating response times was incredibly difficult. To solve this, we designed a lightweight "database snapshot builder" that extracts only critical state changes (expired lots, active alerts, slot occupancies) to keep prompt sizes minimal.
We also spent significant time mastering React 19 state rendering. Updating real-time cold-chain SVG telemetry sparklines every minute for dozens of active zones while maintaining smooth page transitions required intensive profile debugging and memoization tuning to eliminate re-rendering lag.
🏆 Accomplishments that we're proud of
We successfully delivered a fully functional, enterprise-grade logistics console that serves four distinct company roles—Admin, QC, PPIC, and Operator—each with strictly enforced access policies. We are especially proud of implementing a zero-downtime AI fallback mechanism; if the external Roboflow or Gemini APIs are offline, the system seamlessly transitions into a structured manual inspection form, ensuring factory operations never halt. Additionally, we achieved 100% test coverage for critical slot allocation algorithms using property-based testing.
🧠 What we learned
Building AromaSys taught us the power of atomic database transactions. Implementing BEGIN/COMMIT/ROLLBACK across multi-step operations is vital to prevent partial data corruption in rapid industrial environments. We also discovered that property-based testing (using fast-check) is far superior to traditional unit testing at uncovering hidden edge cases in mathematical date-handling algorithms like FIFO expiry calculations.
🚀 What's next?
Our immediate next step is integrating physical ESP32 microcontroller sensors directly with our /api/cold-chain ingestion endpoint to transition from simulated telemetry to real-world industrial hardware. We also plan to train our own Roboflow computer vision models using proprietary extract powder photos to automate fine-grain powder consistency grading, and implement automated SMS/WhatsApp alerts for operators when temperature thresholds are breached.
💻 Built With
• Next.js · React · TypeScript · Tailwind CSS • Express.js · Node.js · Neon Serverless PostgreSQL • Google Gemini 2.5 Flash · Roboflow Vision API • Vitest · Fast-Check (Property-Based Testing) • Vercel · Render · buildpad
Thank you for watching! Leave a comment below with your feedback! 🚀
Built With
- express.js
- gemini
- neon
- next.js
- node.js
- postgresql
- react
- render
- roboflow
- serverless
- tailwind
- typescript
- vercel
- vitest
