How This AI Machine Is Saving Indian Onion Farmers From Rejected Orders and Heavy Losses

“Even after spending so much on grading, we were still losing orders because buyers did not trust the quality,” says Bhanudas Shelke, director of Astitva Agro FPC in Sangamner, his voice reflecting frustration. “One small rejection could wipe out all our profit from a shipment.”

Advertisment

Add The Better India As A Trusted Source

• Stories
• Visual Stories

This is the hardship that thousands of onion farmers in India face every season. Harvesting is labour-intensive, but the real struggle begins after the crop leaves the field. 

Traditionally, onions are sorted and graded by hand. It is a slow, error-prone process that depends entirely on tight labour. Mistakes in sizing or quality assessments lead to losses in revenue, rejected shipments, and even rotting produce. 

Rising wages, intermittent labour shortages, and the subjective nature of grading have made the system unsustainable. For farmers, every season is a gamble, and their hard work often fails to translate into fair earnings.

“We used to take almost two days to ship one order, spending around Rs 20,000 just on grading and sorting,” Shelke recalls. “Even then, buyers would reject our produce. It was heartbreaking and frustrating.”

The problem was evident, but until recently, there was no solution tailored to the Indian farmers. It was a challenge that Kshitij Thakur understood instinctively, whose farming roots gave him a unique perspective.

Growing up on the land

Kshitij grew up in Dighode, Uran, Maharashtra, in a family that farmed rice and mango. From an early age, he learnt the patterns of sowing, watering, and harvesting. But as years passed, he witnessed the struggles of farming under rising costs and unpredictable markets. 

Labour was expensive and unreliable, and post-harvest losses were common. Irrespective of the effort, profits were often swallowed by inefficiencies and a lack of market transparency. Eventually, his family stopped farming, not because the soil failed, but because the system around it failed.

“I saw my parents pour their lives into the farm, and still, the system let them down. I always knew there had to be a better way,” he says.

He pursued mechanical engineering, venturing into industrial automation and computer vision. He worked on AI-powered defect detection systems, smart street-lighting, and even computer-vision-based footwear sizing apps. But his mind kept returning to farms, the ones he knew personally, the ones where technology had barely touched post-harvest processes.

“In factories, defects are obvious, but on farms, quality is a debate. I wanted to bring the precision of AI to agriculture, where it really mattered,” he explains.

Meeting a kindred engineer

Around the same time, Rakesh Barai, an electronics and instrumentation engineer from Gorakhpur, Uttar Pradesh, was developing automation systems for healthcare, fuel extraction from PET waste, and sound fluid dispensing units. 

He was an expert in building machines that could operate reliably under challenging conditions. The duo met at CIBA, a startup incubator in Navi Mumbai, and bonded over a shared interest in solving the problems of Indian farmers.

“We realised that no one was solving the challenges of smallholder farmers,” Kshitij tells The Better India. “We were both working on advanced machines for industries, but the people who needed it most, the farmers, were ignored.”

Together, they wondered if it was possible to build a machine that could quickly and accurately grade crops while handling them with care.

Picking the crop that was the right fit

Their journey began in Mumbai in October 2018, where they founded ‘Agrograde’, a startup aimed at AI-powered agricultural solutions. 

The company later shifted its base to Pune. The first challenge was identifying the most suitable crop for their machine. After careful consideration, they deliberately settled on onions. These crops are notoriously tricky because they vary extensively in size and shape, bruise easily, and peel during rough handling. 

European machines existed at the time, but they were expensive and poorly suited to Indian varieties. Many farmers had tried them only to see their onions damaged, which made them sceptical about adopting new technology.

“Farmers had been burnt before. Convincing them that a machine could work without damaging the onions was the hardest part,” he explains. The team realised that solving this problem would require both mechanical innovation and AI. The handling system had to be gentle, and the grading should be precise.

Building the prototypes

By March 2019, they built their first small-scale prototype. It could sort a few onions at a time, but was slow and inconsistent. The handling system struggled as dust, debris, and the onions’ uneven shapes exposed its weaknesses. From these early tests, the team learnt exactly what needed to improve, setting the stage for larger and more refined machines.

By January 2020, the second prototype addressed some speed and sorting issues, but onions still suffered peeling and bruising. “We realised the biggest challenge was mechanical, not AI,” Rakesh says. “No matter how smart the software was, if the onion got damaged, it was useless.”

The third prototype in June 2020 finally incorporated smoother conveyors, gentle rollers, and early camera-based inspection. With this, they conducted their first pilot with Godadarna FPC in Sinnar, Maharashtra. Farmers could see the machine in action and witnessed onions being sorted without damage.

Kshitij recalls, “When the farmers touched the onions, they said this had not happened before. It was a moment that gave us hope and motivation.”

Between 2021 and 2022, Agrograde developed its fourth, fifth, and sixth prototypes, each iteration faster and more reliable. Pilots were conducted at Pimpalgaon FPC and Umrane FPC in Nashik, testing machines in real-world conditions. Each pilot was free, with farmers providing critical feedback that shaped design decisions.

“We listened more than we spoke. Every suggestion from farmers became part of the machine,” Rakesh says. 

By late 2022, the sixth version was ready for production. It could grade onions at high speed, detect defects with over 96 percent accuracy, and, crucially, handle onions gently without damaging the outer skin.

How the machine works

The machine combines mechanical precision with AI intelligence. Onions are fed into the system, where each passes under an industrial-grade camera capturing multiple angles. The AI model, trained on hundreds of thousands of images collected from farmers across India, analyses size and detects defects such as black smut, rot, sunburn, sprouting, or skin damage.

“The AI is only as good as the data,” Kshitij says. “We spent years collecting samples from different states and seasons. That is what makes grading accurate and trustworthy.”

Once assessed, onions are sorted into categories. This allows farmers to sell high-quality onions at a premium while removing defective ones before storage, reducing losses and standardising quality for buyers.

“For the first time, we could guarantee what was inside every box,” says Praful Bante of Mitraya FPC, Amravati. “Buyers finally stopped questioning our quality.”

The onion grading and sorting machine is delivering results that matter. Shelke reports that Astitva Agro FPC now ships two containers daily instead of one over two days. Grading costs fell from Rs 20,000 to Rs 7,000 per order, and premium grades fetched an additional Re 1 per kilogram.

“Along with saving money, we have credibility now, and buyers trust us. That was impossible before,” he adds.

In Solapur, Balkrishna Patil of Hortimax FPC saw operational costs drop by Rs 0.30 to Rs 0.40 per kilogram. Once grading was standardised, conflicts with farmers faded away. Export markets, once inaccessible, have now opened their doors.

“We can now confidently send containers abroad without worrying about complaints,” Patil explains.

In Amravati, Praful describes a dramatic change. Weekly shipments increased from 10 tonnes to 40 to 50 tonnes, grading expenses fell from Rs 4 per kilogram to Rs 0.5 per kilogram, and 30 to 40 percent of the produce now fetches a premium.

“It feels like we finally belong to the market,” he says. “Middlemen no longer control everything. Technology is giving us a voice.”

Expanding reach

Agrograde now operates across 12 Indian states, working with crops including onions, potatoes, tomatoes, arecanut, and apples. Its presence spans Maharashtra, Pune, Gujarat, Rajasthan, Chandigarh, Jammu and Kashmir, Chhattisgarh, Assam, Andhra Pradesh, Kerala, Tamil Nadu, and Uttar Pradesh. Over 70 units have been deployed, collectively grading 24,000 quintals per day. 

The company works directly with eleven Farmer Producer Companies, including Astitva Agro FPC in Sangamner, Hortimax FPC in Solapur, Mitraya FPC in Amravati, Godadarna FPC in Sinnar, Pimpalgaon FPC in Nashik, and Umrane FPC in Nashik, to standardise quality and access new markets. 

Post-harvest losses are reduced by up to 90 percent, costs are lowered dramatically, and revenue increases by around 10 percent per kilogram. Rental models make the technology accessible even for smaller FPOs.

“We did not want to sell machines to a few rich farmers. We wanted every farmer to benefit, even indirectly,” Kshitij says.

Recognition, but grounded in purpose

Agrograde has been recognised nationally and internationally. Awards include Top 9 Agritech Startups by Mahindra Startup Leap in 2023, recognition in NITI Aayog’s 75 Agri Entrepreneurs, first national runner-up in MANAGE Samunnati Agri Startup Awards in 2022, BSE Top 10 Impact Ventures in 2019, and memberships in programmes by NVIDIA Inception, BOSCH DNA Accelerator, Social Alpha, Villgro, ICAR DOGR, and CIBA.

“Awards are nice, but adoption is everything. If a farmer benefits, that is the real recognition,” Kshitij adds. 

For the founders, especially Kshitij, Agrograde is close to the heart.

“I could not save my family’s farm. But helping thousands of others survive and grow feels like fulfilling the promise I made to myself in my village,” he explains.

For the farmers, the machine is life-changing. “Technology gave us back control over our crops. We are no longer at the mercy of middlemen or luck. That is priceless,” Bante says.

The journey from childhood fields in Uran to a nationwide supply chain revolution is a story of patience, humility, and consistency. It took six prototypes, countless farmer conversations, thousands of images, and years of mechanical and software innovation to create a solution that feels natural to farmers.

“We learnt humility from the farmers,” Kshitij says. “Every time we thought we solved a problem, they showed us there was more work to do.” Today, the startup stands as proof that technology, when designed with empathy and experience, can create fairness and efficiency in agriculture.

All pictures courtesy Kshitij Thakur.