Jhutlink

Bangladesh produces 6 lac tons textile waste annually. Jhutlink is an AI native B2B platform that turns that waste into opportunity by connecting suppliers & buyers in a circular economy marketplace

Comment

Jhutlink

Inspiration

Bangladesh is the world's second-largest apparel exporter, yet every year the industry generates nearly 600,000 tons of textile waste. Despite being rich in reusable cotton, yarn, and fabric scraps, most of this waste is traded through fragmented informal networks, exported without value addition, or discarded inefficiently.

This is more than an environmental problem.

It is a lost economic opportunity.

Industry experts estimate that Bangladesh could unlock billions of dollars in value from textile waste while reducing raw material imports and creating new jobs. Yet there is no unified system to identify, classify, trace, and trade textile waste at scale.

We built Jhutlink to become that system.

What It Does

Jhutlink is an AI-native B2B platform that transforms textile waste into a traceable, tradable asset.

Manufacturers can upload waste inventory, AI automatically classifies and categorizes materials, and buyers — such as recyclers, spinning mills, textile processors, and circular manufacturers — can discover and purchase suitable materials.

Jhutlink creates transparency in a market that currently operates largely through disconnected intermediaries and informal networks.

How We Built It

Architecture

The frontend is a React 19 + TypeScript SPA with Vite, Tailwind CSS, and shadcn/ui, deployed as a static site on Vercel. The backend is a modular FastAPI server with 12 single-responsibility services (auction engine, escrow state machine, matching engine, order manager, AI service, etc.), entity models with finite-state-machine validation, and JSON file persistence ready for PostgreSQL migration. The entire platform runs on a single Python process with background sweep threads for auction expiry and payment timeouts.

Tech Stack

  • Frontend: React 19, TypeScript, Vite, Tailwind CSS, shadcn/ui (53 Radix components), GSAP + ScrollTrigger + Lenis for cinematic scroll animations, Recharts for data visualization
  • Backend: FastAPI (Python 3.11), uvicorn, JWT auth (python-jose + bcrypt), WebSocket for real-time bid updates
  • AI: OpenAI GPT-4o-mini via openai Python SDK — generates structured market intelligence (price forecasts, district spotlights, pattern detection) from live platform data with 5-minute caching
  • Deployment: Vercel (frontend), GitHub (source: github.com/astra-builds/Jhutlink)

OpenAI Integration

The AIService collects real marketplace data — listings, orders, bids, user activity, price spreads, grade distributions, district-level pricing — and sends it to GPT-4o-mini with a structured JSON prompt. The model returns five fields:

  1. A market summary
  2. Per-waste-type price forecasts with direction, range, and reason
  3. A district spotlight
  4. A key actionable insight
  5. Detected patterns and anomalies with confidence and impact ratings

When no API key is configured, the service falls back to rule-based insights (price gap analysis, bid match rate evaluation, spread detection). Cost is ~$0.0002 per call.

Challenges We Ran Into

The biggest challenge was that textile waste is not standardized.

Different factories use different naming systems, grading methods, and trading practices. Much of the industry's waste movement is also difficult to track due to fragmented supply chains and informal transactions.

Building a system that could bring structure, transparency, and trust to this ecosystem required us to rethink how waste is identified, valued, and exchanged.

Accomplishments That We're Proud Of

Jhutlink addresses a problem that sits at the intersection of:

  • Economic growth
  • Environmental sustainability
  • Industrial innovation
  • Employment generation

By digitizing textile waste flows, we are helping transform a neglected byproduct into a measurable economic resource.

What excites us most is that the opportunity is already proven. Bangladesh produces enough textile waste to support a large-scale circular economy — what has been missing is the infrastructure to connect all stakeholders.

What We Learned

We learned that waste is often a data problem before it becomes a recycling problem.

The challenge is not the absence of valuable materials. The challenge is the absence of visibility, traceability, and coordination.

We also learned that AI can play a critical role in circular economies by helping industries identify, classify, and route materials more efficiently.

What's Next for Jhutlink

Our vision is to become the operating system for textile waste management in Bangladesh.

In the future, Jhutlink will:

  • Connect factories, recyclers, traders, and manufacturers nationwide
  • Enable traceable textile recycling supply chains
  • Support ESG and sustainability reporting
  • Provide waste analytics and demand forecasting
  • Integrate logistics and procurement workflows

If Bangladesh can capture even a fraction of the value currently being lost from textile waste, the impact could be transformative: lower import dependence, stronger recycling industries, thousands of new jobs, and a more sustainable apparel sector. Experts estimate that local recycling alone could save hundreds of millions of dollars annually, while a mature circular ecosystem could unlock billions in economic value.

Jhutlink isn't just helping factories sell waste. It is helping Bangladesh build the infrastructure for a multi-billion-dollar circular textile economy.

Built With

  • fastapi
  • openai
  • python
  • shadcn
  • tailwind
  • vite
Share this project:

Updates

Submission history