Mr. Kshitij Thakur | Co-Founder & CEO | Agrograde

Q1. What inspired you to start Agrograde, and what gap in India’s fresh produce supply chain did you set out to solve?

Answer: I previously worked in industrial automation, where advanced technologies were used to assess quality, improve efficiency, and bring transparency to manufacturing. When we looked at fresh-produce supply chains, we saw a major gap: many operations still relied on manual work, quality assessment was subjective, and suitable technology was often too expensive for much of the Indian market. This created disputes between buyers and sellers and made it difficult to standardise quality.

We believed that technologies from industrial automation could be adapted for fresh produce. The opportunity to help farmers get fairer prices, reduce post-harvest losses, and improve transparency across the supply chain motivated us to start Agrograde.

Q2. Could you take us through your journey from the early idea to building Agrograde into a company serving customers across multiple states?

Answer: We began working on the idea in 2018 with a mobile application that used images of produce samples to generate quality reports. We soon realised that identifying quality issues was not enough; produce needed to be graded and standardised before dispatch. That led us to develop machines.

We initially explored a machine for multiple commodities, but each crop required a different handling approach. We focused first on onions. In 2020, we piloted a 1 MT-per-hour machine with Godadarna Farmer Producer Company in Nashik. We then tested a 3 MT-per-hour version in Pimpalgaon and went through around six versions before refining the mechanism. We began working on potatoes in 2022 and later expanded to apples and arecanut. Today, we have nearly 150 machines across 15 states.

Q3. Agrograde uses AI, computer vision, and robotics for grading and sorting. How does the technology work in simple terms, and what makes it different from manual methods?

Answer: Our Vector and Inspector series use AI to assess the grade and quality of fresh produce. Produce enters through a hopper and moves through an inspection section. Depending on the crop, it is rotated so industrial cameras can capture multiple views. AI models analyse the images for quality characteristics and external defects, and the sorting mechanism directs the produce into the right output category.

For example, our arecanut system captures around 10–12 images of each nut so it can assess several sides. Manual sorting depends on each worker’s experience and interpretation, which can vary over time and at higher volumes. Automation helps customers apply a defined grading standard more consistently. For simpler size separation, mechanical graders may be the better choice.

Q4. What were the biggest technical and on-ground challenges in building machines suited to Indian crops, mixed produce lots, dust, and varied operating conditions?

Answer: Produce varies greatly in size, shape, and condition. A mechanism that works for medium-sized onions may struggle with very large or small onions, flat shapes, or long stems and roots. We spent considerable time studying how onions roll, how their outer skin can be affected, and what happens when they bounce or fall from different heights. These observations directly shaped our machine design.

Our machines also need to operate in packhouses and open-air settings with dust, humidity, and changing temperatures. They must be easy to operate and maintain. Field trials were essential because many of these challenges do not appear in controlled factory demonstrations.

Q5. Agrograde’s machines serve onion, potato, apple, and arecanut supply chains. How do you adapt the technology to the different characteristics and grading needs of each crop?

Answer: We begin by understanding the crop and how its quality is assessed in the supply chain. Onions have loose outer skin, so excessive friction can affect their appearance and storage quality. Potatoes may have soil and mud on their surfaces, while apples need gentle handling to avoid bruising.

Arecanut is smaller, and defects cannot always be identified by colour alone. The machine must present and rotate each nut correctly so cameras can capture the details needed for inspection. We adapt the feeding mechanism, produce movement, imaging arrangement, sorting mechanism, and AI models for each crop. Experience helps us develop faster, but every crop still needs to be studied carefully.

Q6. The company has deployed more than 150 machines across 15 states. What have been the key lessons from scaling deployments and earning customers’ trust?

Answer: We are now at nearly 150 machines across 15 states. One key lesson is that customers need time to trust a new technology that will become part of their daily operations. They want to know whether it works across locations and seasons, how long it will last, how easily it can be maintained, and whether support will be available after the sale.

Customers have seen our machines evolve through different versions and have seen how we respond to problems during commercial use. That continuity has helped us build trust. India is also a diverse market, so crop varieties, operating practices, and customer expectations differ by region. We must keep listening and adapting.

Q7. How do Agrograde’s solutions help customers improve grading consistency, reduce handling losses, and make post-harvest operations more efficient?

Answer: Our technologies have made many packhouse processes autonomous. They help customers prepare a defined grade more consistently, giving buyers greater confidence and reducing quality-related disputes. Customers can also plan processing and dispatches more reliably.

We pay close attention to transfer points, impacts, and how produce moves through a machine to reduce avoidable damage. Since fresh produce has a limited shelf life, reducing delays caused by grading and labour shortages can help customers dispatch earlier. The benefits include more efficient operations and greater reliability across the process.

Q8. How does Agrograde’s business model work, and how do you make automation commercially viable for FPOs, aggregators, traders, exporters, and packhouses?

Answer: Our main business model is selling equipment to businesses that aggregate and handle produce, including FPOs, traders, exporters, and packhouses. To assess whether automation makes commercial sense, we look at daily volumes, operating months, manual grading costs, quality-related losses, dispatch delays, and whether customers can separate grades that different markets value differently.

Customers who mainly need size grading may use our GradePlus series, while those who need external quality inspection and defect separation require optical sorting. We focus on making our machines affordable, easy to operate, and straightforward to maintain.

Q9. Can you share an example of how Agrograde’s technology has changed a customer’s day-to-day operations or improved the way produce is graded and traded?

Answer: One example is Mitraya FPC in Amravati, which handles white onions. Earlier, preparing a 10-tonne consignment took two to three days and primary processing cost around ₹3 per kg. After adopting our machine, the same consignment could be prepared in around three hours with substantially fewer people. Separating a premium size grade also helped them approach buyers who paid an additional 50 paise to ₹1 per kg for that portion.

Hortimax FPC in Solapur used our GradePlus series to separate an onion grade that attracted export buyers willing to pay around ₹2–₹2.5 more per kg. At Raj Traders in Mangalore, our Vector Series AI optical sorting system helps process close to eight tonnes of arecanut per day, with capacity of up to around one tonne per hour. These examples show how consistent grading can improve planning, quality control, and access to markets.

Q10. What were the most difficult moments in building the company, and how did you navigate setbacks while moving from product development to commercial deployment?

Answer: There were periods when we struggled to pay salaries while funding the next machine version. Sales were slow, products sometimes failed during demonstrations, and we had to rebuild customer confidence. We also lost talented team members because we could not afford to retain them. In hardware, a prototype is only the beginning; the product must then become reliable at customer sites and suitable for consistent manufacturing at scale.

Our team made sacrifices to help the company continue, and our early customers, mentors, and investors were important. Aligned Ventures backed us early, while Social Alpha, Villgro, and IIMA Ventures supported us during the pilot stage and beyond. Incubators and accelerator programmes also helped. Over time, we created an internal guidebook based on field experience to reduce repeated mistakes and improve our development process.

Q11. How do you collect and use field data and customer feedback to improve your AI models, machine performance, and grading accuracy over time?

Answer: Our team collects data across geographies, seasons, and crops, while working closely with farmers, traders, and customers. We study how experienced people assess quality, what causes customer complaints, why consignments receive debit notes, and which defects affect prices. Field images help improve our AI models, while customer feedback can also lead to hardware changes if the issue relates to feeding, rotation, or how produce is presented to the cameras.

Our work with Raj Traders in Mangalore is one example. Their team helped us understand arecanut quality requirements and real operating conditions. Continued use after installation gives us insights into crop variations and challenges that are difficult to reproduce fully during development.

Q12. What sets Agrograde apart from conventional grading equipment and other automation solutions, and how do you continue to build a lasting competitive advantage?

Answer: Our differentiation comes from understanding each crop in depth before designing a machine. For onions, that meant studying how they roll and lose their skin, how traders assess quality, and why buyers reject consignments. Conventional size graders address a specific need, while optical sorting can assess external quality and defects. We develop solutions based on what the customer actually requires.

We handle design, manufacturing, and software development in-house, allowing us to improve the full system. Mechanical handling, imaging, software, and sorting need to work together; AI alone cannot solve every problem. Our patent-pending technologies, field data, engineering experience, customer relationships, and after-sales service all contribute to our long-term advantage.

Q13. As you develop solutions for crops such as tomato, garlic, and leafy vegetables, how do you decide which new crop categories or markets to enter next?

Answer: New crop development often begins with customer enquiries. We assess how serious the problem is, how many businesses face it, whether the solution can serve more than one customer, and how technically feasible it is. We also consider the development effort required and whether we have a partner who can validate the machine in real operating conditions.

Our arecanut development followed this approach: the customer brought knowledge of the crop and trade, while we worked on engineering and AI. Although our process has become faster, some crops still require substantial research. Given our capacity, we can take on only a limited number of new crops each year and want to ensure each solution becomes reliable and commercially viable.

Q14. What role do your team, manufacturing capabilities, and partnerships play in scaling the business, and what capabilities will be most important for the next stage of growth?

Answer: Our team is central to our progress. Building these machines requires people who can observe a practical problem, develop a mechanism, build it, and improve it through testing. We are strengthening our design and engineering capabilities, especially in hardware, and value hands-on experience in taking concepts through to execution.

In-house design, manufacturing, and software teams can respond closely to field feedback. As deployments grow, manufacturing quality and service capacity must grow too. Pilot customers help us understand crop behaviour and trade needs, while channel partners support customers and share regional insights. Our next stage requires us to develop new products while reliably manufacturing, installing, and supporting existing machines across more locations.

Q15. Looking ahead, what is your long-term vision for Agrograde and the autonomous packhouse, and how do you see technology reshaping India’s post-harvest ecosystem?

Answer: Our long-term vision is to build autonomous packhouses where handling, quality assessment, grading, sorting, and packing require very little manual intervention. We see this as an important infrastructure layer for fresh-produce supply chains. Consistent quality assessment can give buyers and sellers a stronger basis for trade and support faster market linkages, greater transparency, and more confidence in transactions.

This infrastructure could also support digital commerce in fresh produce and help more businesses access export markets. We are starting with grading and sorting and gradually working towards automating more of the packhouse. We believe India can become an important source of agricultural machines, robotics, and technology, and we want Agrograde to be among the leading companies building from India for the world.

Interview By: Palak Pandey