Second Harvest

Inspiration

We kept noticing the same contradiction everywhere: packaged food, still sealed and safe, being pulled off retail shelves and thrown away, while families a few streets away struggled to afford those same branded essentials. India wastes 78 million tonnes of food every year, worth over ₹1 trillion, and a large, structurally invisible share of it comes from organised retail and FMCG supply chains, not farms or households. Shelf compliance rules force products with fewer than 30 days of shelf life remaining to be discarded, even though they're nowhere close to expiry. That gap between destroyed value and unmet need is what we set out to close.

What it does

Second Harvest is an AI-powered reverse supply chain platform that intercepts near-expiry, returned, and excess packaged food from retailers, manufacturers, warehouses, and dark stores before it's destroyed. Machine learning models monitor SKU-level inventory in real time and predict which products are likely to be written off, flagging them up to 14 days before disposal thresholds hit. Recovered inventory is quality-validated against FSSAI norms, graded, and redistributed through a consumer app, community resellers, and hyperlocal WhatsApp commerce at 30 to 70% below MRP. Anything unfit for human consumption is converted into animal feed, so nothing reaches a landfill.

How we built it

We started by mapping the structural cause of the problem: surveying shelf compliance practices at organised retailers and quick commerce operators, and identifying that no technology-enabled recovery infrastructure exists for packaged food in India, unlike fresh produce or cooked food donation models. We designed a five-stage operating methodology, data intelligence, procurement, sorting and quality validation, redistribution, and residual processing, and built out the unit economics behind it. To ground the model in reality, we ran small scale prototype testing with local ration shop owners, validating sourcing, demand, and redistribution feasibility directly with the retailers and households it's meant to serve.

Challenges we ran into

The biggest challenge was designing a model that works at the intersection of two very different worlds: technology systems and physical, on-the-ground logistics. A platform that's technically strong but operationally fragile won't survive contact with a warehouse floor, and one that's operationally strong but lacks intelligent prediction won't scale. We also had to navigate consumer trust around near-expiry food, since "near-expiry" often gets wrongly read as "near-unsafe," which meant building transparent labelling and FSSAI-aligned grading into the core of the product, not as an afterthought.

Accomplishments that we're proud of

We've built a fully costed, execution-ready 30-day hyperlocal pilot design, validated demand and feasibility through direct prototype testing with ration shop owners, and had our concept recognized by premier institutions across India. We also identified concrete supply chain partners and mapped a path from a single neighbourhood pilot to a pan-India recovery network.

What we learned

We learned that the food itself isn't the problem, the supply chain's thresholds are. Compliance rules built for safety end up discarding perfectly good food because there's no infrastructure to catch it in between. We also learned how much trust matters at the last mile: recovery only works if local resellers and households believe in the safety and value of what's being offered, which is why community-rooted distribution matters as much as the AI layer.

What's next for Second Harvest

Our immediate next step is running the 30-day hyperlocal pilot in one neighbourhood, partnering with local ration shop owners and nearby retailers to recover near-expiry inventory and redistribute it to bottom-of-the-pyramid households. From there, we aim to expand city by city, deepen our AI recovery engine, and engage FSSAI and the Ministry of Food Processing Industries to help shape a near-expiry redistribution policy framework that doesn't currently exist in India.

Built With

• ai-inventory-prediction
• business
• fssai-compliance-frameworks
• machine-learning
• node.js
• python
• react
• whatsapp