SupplAI

The description states that SupplAI is a platform that:

• Turns market signals into ranked, explained supply decisions
• Provides demand intelligence through interactive charts with confidence bands
• Offers scenario simulation using a constrained optimiser (OR-Tools)
• Includes decision workflow with approval/rejection capabilities
• Processes free-text market news into structured signals that affect planning
• Allows data import from CSV with validate/commit flow

The platform is described as having:

• A web interface built with Angular 19
• An API layer using Express 5 + Zod validation + JWT/RBAC
• An optimiser service built with FastAPI + Google OR-Tools (CBC)
• An AI layer that provides explanations and news→signal extraction
• Data storage via PostgreSQL through Prisma

The system is described as fully functional end-to-end, with a runnable vertical slice including login → forecast → optimise → explain → decide → audit.

Evidence strength Self-reported. No independent verification of functionality or performance.

The description states SupplAI's positioning:

• It is not another dashboard that describes what happened
• It is a prescriptive decision co-pilot that senses demand, reads the market, and tells planners exactly what to do
• It provides math-backed recommendations with human control over commitments
• It uses AI to explain decisions and structure news, but never invents numbers

The claim evolution shows:

• Initial inspiration: petrochemical producers need better allocation decisions than spreadsheets and gut feel
• Core value proposition: prescriptive decision-making that combines forecasting, optimisation and explanation
• Key differentiator: human-in-the-loop approach where AI explains but doesn't compute allocations

Evidence strength Self-reported claims about positioning and evolution.

The description states:

• The primary customer is petrochemical producers (specifically those selling PE and PP grades)
• These producers operate in Southeast Asia markets
• The target user is supply planners who make daily allocation decisions
• The platform is designed for non-technical planners who can actually read the command centre

Evidence strength Self-reported customer positioning.

Not evidenced. No information provided about pricing, licensing, or revenue model.

The description states:

• Built with Angular 19 (standalone components, Signals, strict TypeScript)
• API built with Express 5 + Zod validation + JWT/RBAC
• Optimiser service using FastAPI + Google OR-Tools (CBC)
• AI layer using Claude Opus 4-8 or deterministic fallback
• Data storage via PostgreSQL through Prisma
• Platform works with zero API keys
• End-to-end vertical slice verified in real browser and against live Postgres + OR-Tools
• 69 passing API tests, clean production build, one-command docker compose up

Evidence strength Self-reported technical implementation details.

Not evidenced. No information provided about:

• Revenue or customers
• Product usage metrics
• Market adoption
• Growth trajectory
• Any commercial activity beyond the hackathon submission

Not evidenced. No information provided about:

• Competitors in the supply planning space
• Market size or competitive landscape
• Differentiation from existing solutions

Inferences based on self-reported evidence:

1. Unproven commercial viability: The project is described as a hackathon submission with no evidence of traction, revenue or customers.
2. Limited team size: Only one team member (alson-how Elson) is mentioned, raising questions about execution capability.
3. No external validation: No third-party reviews, user feedback or pilot data are provided.
4. AI boundary discipline: While described as a key feature, it's unclear how this would translate to real-world reliability and trust at scale.
5. Data quality assumptions: The platform relies on importing real historical data for forecasting, but no evidence of data sourcing or cleaning processes is provided.

Evidence strength Inferences from self-reported information.

1. What specific margin improvements have been observed in pilot testing against current spreadsheet methods?
2. How does the platform handle integration with existing ERP systems used by petrochemical producers?
3. What is the timeline for moving beyond the hackathon prototype to a production-ready solution?
4. How do you plan to scale the AI layer to handle more complex scenarios without losing deterministic control?
5. What are the key assumptions in your demand forecasting model, and how have they been validated?
6. Can you demonstrate actual performance metrics from the constrained optimiser compared to manual planning?
7. How will you ensure data quality when importing from CSV sources with potential formatting issues?
8. What is your go-to-market strategy for reaching petrochemical producers in Southeast Asia?