How this Gurugram-based startup turns e-commerce sites into a visual treat

Spyne, based on Gurugram, helps e-commerce platforms create and scale high-quality product images at scale with AI. Founded in 2018, the tech startup serves over 80 enterprise customers in over 15 countries and over 23 markets, processing over 500 images per minute on average.

“In general, investment is looking to dry up in the tech space. But, in my view, good companies with strong financials and growth prospects will always get capital from customers and investors. In fact, in this environment, companies with strong business models and low burn rates would take a big leap forward and take the next step,” said Spyne CEO Sanjay Kumar.

In an exclusive interaction with Analytics India Magazine, Sanjay discussed Spyne’s journey so far.

AIM: What prompted you to create Spyne?

Sanjay: I was in the tech industry for almost 13 years before founding Spyne. After working mainly with companies based on e-commerce platforms, I realized that cataloging is a challenge that most companies face. Sellers come to these platforms with a passion for selling their products online. Seeing cataloging issues up close, I realized that this is one of the fundamental problems of any online service.

With this idea we started four years ago, I didn’t want to create a cataloging platform and instead made a content site for digital creators and photographers initially. However, after a year of activity, we felt that the current model was not scalable. Photographers and digital creators don’t necessarily have deep pockets, and providing a SaaS platform to such a niche market has proven unfeasible.

Our goal was to help create an effective version of photography. We started with a different business model two and a half years ago, similar to Nero.ai but with more emphasis on automation. We ventured into the automotive space in the first year of operation, which picked up well.

AIM: How does Spyne’s AI-driven automated image processing engine work?

Sanjay: I would like to start with the heart of why and how this engine was built. Companies like Myntra, OYO, Zomato, Swiggy, etc. cover a huge market with a uniform cataloging process. The products on these platforms look very presentable. For example, if you see a shoe catalog, the products look very appealing because these companies have created their own guidelines on how the products should be displayed.

Armed with knowledge of these guidelines, one can build a system capable of automating many of these rules, which formed the premise on which we began to build an automated computer vision model. The template has solved some of these issues one at a time like background removal, shadow generation, reflections, body color correction, and more.

The model also addressed some creases or deformations on the product. After selecting about 20-30 such issues, we created an efficient data pipeline and set up a core team to create specialized data. We also collaborated with a large team of freelancers who worked on our data pipelines in general to dig deeper into each category.

After determining the top-scoring issues in each category, we began creating data and training our model. Some of the research was already available in articles, methodologies and repositories. We took some concepts from industry, open source publications and started to put all this information together in a model answering specific problems.

We have additionally created many GAN-based templates to turn raw footage into studio-quality finished footage. We perform many image corrections such as removing background, adding background, generating realistic shadows, cleaning windows, cleaning body reflection, correcting light, etc Each problem was solved by running GAN models on millions of custom datasets.

Because every problem requires specialized data, we purchased and generated high-volume image data to continuously build and refine existing models. This helps us improve the accuracy of existing solutions in a very agile way.

AIM: What is the biggest challenge Spyne has had to face?

Sanjay: The hardest issue we faced came early on as we struggled to identify the issue, which was not letting our AI engine deliver accurate results. We realized that there were several issues, which required focused resolution and data training to increase the accuracy of our AI models.

For example, how to bring out the best image of a car if the car is not taken from the top angle or the light is too strong when the car is taken in daylight. There was also the issue of low light if shot during the evening, environmental objects or trees showing through transparent windows, building/tree reflections coming off the car body, etc.

These were real challenges, and getting the training data for such issues was extremely difficult. Eventually, we set up a specific data annotation team for this purpose, which helped us generate targeted data for each of these issues from our current client projects. This strategy helped us improvise and train our AI/ML models to make them more efficient and accurate over time. As a result, our models have improved their outputs from around 60-70% accuracy at the start to over 99% accuracy.

AIM: How has your journey been so far?

Sanjay: It took a long time to set up this particular combination, but the idea was still there to set up an online platform. It made sense to put something online in the early days of Yatra; but things didn’t work out, and I couldn’t take the plunge initially. This decision to wait was a godsend because I acquired interesting knowledge by joining a large company, where one gets to work on a small problem in a very large configuration.

Normally, when working with startups, one is exposed to a wide variety of issues that give them a unique understanding of the nature of the business. Exposure to these issues, strategies and the market scene gave me the experience to start Spyne. Everything I dealt with from the beginning laid the foundation for my idea.

AIM: What do you think of the big resignation? What is Spyne’s strategy to fight attrition?

Sanjay: 2021 has been a great year for tech companies. A wave of massive digitization has swept the world triggered by the COVID-19 pandemic. There has been a great influx of capital into the market, helping companies create solutions for the future. So we saw a high demand for good tech talent across the board, leading to high employee attrition from organizations that couldn’t engage well enough with the best.

At Spyne, we were able to create a rising high-tech SaaS brand, creating unique AI solutions that would revolutionize the very traditional e-commerce photography industry. We are not hiring like hundreds, but selective talents, who believe in our big vision and want to be part of this transformation.

Over the past year, we have seen near zero attrition. We have been able to retain our talents due to the nature of our work, the friendly work environment, the positive work culture and the growth path we have provided for our business. We have been able to effectively communicate our company’s vision and empower everyone associated with us.
The timing was right for a company like ours, a start-up SaaS company with low utilization. We just raised enough funds to build our growth. We will continue to focus on building strong fundamentals and creating product and value for every part of the ecosystem.

Comments are closed.