How an Ahmedabad Student Built an AI Waste Sorting Startup

How an Ahmedabad Student Built an AI Waste Sorting Startup

With a ₹70,000 student grant and a smoking prototype in a college lab, SegReClaim is trying to fix India’s trash crisis before garbage hits the landfill.

At 2:15 AM in a cluttered garage workshop, the smell of burnt plastic filled the air. A single crushed soda can had jammed the mechanical trapdoor of a metallic box, causing a low-cost stepper motor to short-circuit the microcontroller board below it. Wiring sparked, smoke billowed, and the camera went dark. Sitting on the concrete floor, Harshil Patel wiped grease off his hands and looked at his phone. His college WhatsApp group was buzzing with corporate job offers. This is the story of how an Ahmedabad student built an AI waste sorting startup while his peers chased IT salaries.

"Everyone in my batch was preparing for coding interviews and discussing 12-LPA packages," Patel recalls, leaning back against a workbench piled high with circuit boards. "I was sitting alone in a dusty room with a burnt motor, ₹4,000 left in my bank account, and a machine that couldn't even sort a cola can without catching fire. I remember asking myself, 'What am I doing with my life?'"

That night was the closest Patel came to packing up his tools and surrendering to a standard corporate job. Instead, he stayed.

To understand why someone would choose burnt wiring over a comfortable corporate career, you have to look at what happens every evening on the outskirts of Gujarat's commercial hub. A few months earlier, Patel—then in the final year of his B.Tech in Computer Engineering—visited local municipal dump yards and regional waste-to-energy facilities. What he saw was a systemic disaster. Heavy dump trucks arrived every six minutes, tipping mountains of wet kitchen scrap, shattered glass, corrugated cardboard, and plastic bottles into decaying piles. At the waste-to-energy plants, expensive processing machinery ground to a halt for hours because unsegregated metal cans jammed the feeder blades.

India generates over 62 million tons of municipal solid waste every year, but less than 20 percent of it is segregated at source. Municipalities pour hundreds of crores into setting up manual sorting yards at landfills. By the time mixed trash sits under the sun for forty-eight hours, organic juices seep into paper, glass shatters into plastic, and high-value recyclable materials become worthless contaminated sludge.

"The whole system is built backwards," Patel says. "We spend billions trying to clean up mixed garbage after it has already rotted in a truck. The battle is lost in the first three seconds—the exact moment someone drops an item into a public trash bin."

Most clean-tech founders try to solve this by building massive sorting plants at landfills or launching public awareness campaigns to educate citizens. Patel knew both approaches were flawed. You cannot rely on human behavior when a hurried commuter is dropping a wrapper on their way to catch a bus. And you cannot economically un-mix trash once it hits a dump truck.

The hidden insight was simple: the sorting had to happen automatically, mechanically, and silently at the drop-off point.

In early 2026, Patel teamed up with his classmate Rikin Parekh to launch SegReClaim. Their goal was to build a heavy, automated smart bin powered by edge computing and computer vision. When a person approaches the bin and drops an item into the slot, an internal camera captures high-speed images. An onboard microcontroller runs custom computer vision models, using majority-proportion classification logic to evaluate mixed or single-item deposits. Within milliseconds, mechanical actuators open specific internal trapdoors, routing plastic, glass, paper, and metal into separate clean chambers while logging the deposit to reward the user.

Building that three-second physical interaction, however, exposed the brutal reality of hardware engineering.

When the founders tried transitioning their initial mechanical frame into an automated edge-AI system, everything failed simultaneously. Direct sunlight entering the drop slot blinded the low-cost camera lenses. Wrinkled foil wrappers confused the image classification models. Low-voltage microcontrollers froze under the real-time processing load, and heavy mechanical actuators jammed whenever a user dropped multiple items at once.

"Software developers can push a hotfix at midnight and fix a bug," says Patel. "When hardware fails, sheet metal bends, motors burn out, and you have to spend three days waiting for a replacement part to arrive from a supplier. There were weeks when we doubted whether low-cost microcontrollers could ever handle real-time edge classification in humid, dusty Indian street conditions."

Clean-tech hardware in India is an unglamorous, exhausting grind. There are no cloud credits or high-margin software subscriptions to shield you. You get sharp metal cuts, fried circuit boards, zero revenue, and polite rejections from risk-averse investors who want to fund the next food delivery app instead of a physical machine that handles garbage.

The turning point arrived on August 15, 2025. SegReClaim was awarded a ₹70,000 grant under Gujarat’s Student Startup and Innovation Policy (SSIP).

It was not a mega venture capital round, but for two broke students, it was everything. In August 2025, SegReClaim secured ₹70,000 in grant funding from the Student Startup and Innovation Policy to build its first working AI smart bin prototype. Every single rupee was spent with surgical precision. They bought industrial-grade microcontrollers, upgraded their camera modules, and fabricated a reinforced sheet-metal casing that could withstand public use.

Operating as a pre-revenue startup with a team of fewer than ten people, they channeled every resource into building one fully functional prototype. But a laboratory prototype means nothing until real people use it.

To test the machine under real-world conditions, Patel and Parekh hauled their heavy smart bin to the central corridors of Indus University. Students gathered around between lectures, dropping empty plastic bottles, discarded notebook pages, and aluminum soda cans into the slot. The bin paused for a fraction of a second, an internal camera flashed, the edge-AI model classified the item, and a crisp mechanical click echoed as the correct internal chute opened. The machine worked cleanly.

Following the success at Indus University, the team moved the prototype to IIT Gandhinagar for secondary pilot testing and showcased its live automated sorting during the Shashwat Urja National Conference. Across these trials, the machine proved that source segregation could be completely automated at the point of drop.

Everyone tells founders to raise venture capital quickly. SegReClaim built a working edge-AI hardware prototype on a student grant before taking a single rupee of equity funding—and that is why they still retain total control of their company.

Yet, solving the engineering puzzle revealed an even steeper wall: the Indian state. Today, while SegReClaim’s machine functions reliably in campus pilots, their primary bottleneck is navigating government bureaucracy. Moving smart bins from university halls onto city streets requires navigating municipal offices, administrative approval chains, and slow public tender cycles.

"I used to think designing neural networks on edge devices was hard," Patel says with a quiet laugh. "Then I had to sit outside a municipal officer's cabinet for four hours just to get a ten-minute meeting about street infrastructure permissions. Bureaucracy moves at its own speed, and as a young founder, learning to survive that delay without running out of momentum is the hardest lesson of all."

SegReClaim remains a pre-seed startup operating with one working machine, zero institutional equity funding, and a massive challenge ahead. Patel and Parekh are not taking victory laps. They know that a successful campus pilot is just a preliminary step toward fixing a nationwide crisis.

Tomorrow morning, Patel will pack his laptop, grab a tablet showing live video of his smart bin sorting waste at IIT Gandhinagar, and head to another government office in Ahmedabad. The machine is built. The AI works. Now, he just has to convince the city to let it clean the streets.