Does Fashion eCommerce Industry Need a Dose of AI to Evolve Itself? Indian Startup Streamoid Thinks So


Shown above – Team Streamoid

Even though Fashion is the most popular category shopped online, it’s still a hard game to crack. Fashion eCommerce companies always have an opportunity to up sell to a customer they have acquired, the tough part is to figure out ‘how’.

Personalisation & gamification are now two commonly used techniques by these firms. It’s impossible to miss the ‘You may also like’ recommendations that show up everywhere once you buy or just visit an online retailer. These attempts at personalisation are usually not so accurate. Globally technologies like Artificial Intelligence, Augmented Reality, Virtual Reality etc.are being deployed to eliminate the difference between the tangible shopping experience and the digital one.

Companies may boast the size of their catalogues to lure customers. However, bigger the catalogue, bigger the task for a visitor to sift through it, and an even bigger the challenge to create an accurate recommendation engine for each user. Bangalore-based Streamoid, believes that these issues can be collectively resolved by bringing AI into the picture, and is building products to do the same.

A Background into the Idea

Streamoid was started in 2013 by Sridhar Manthani and Rajesh Kumar. The duo and its initial team was involved in deep research in image analysis using computer vision. They were certain that visual content was going to be pervasive across apps and websites.

“With our proprietary research in hand, we set about thinking of how information available in images can be used to improve the customer experience. We decided to focus on fashion eCommerce which primarily uses images to engage and inspire their audiences.” says Rajesh Kumar, CTO, Streamoid.

He further adds “Product discovery in Fashion is a daunting task because of numerous choices. It is difficult to filter what is available based on personal preferences which are hard to describe textually. We are building AI algorithms that help fashion retailers improve their business in several areas. Our first products focused on Improving conversions and cross-sells and reducing the cost of cataloguing.”

They are attempting to mimic the way a personal stylist would recommend a product for a customer. To do this they are using multiple methods and technologies such as an image recognition, Fashion style rules engine, latest trends on social media,  Deep learning, Natural language processing etc.

The Core Product


Streamoid’s piQit Fashion is an AI platform focussed on improving personalisation in the fashion industry. They are building technology that can act as a virtual shopping assistant for shoppers, and is able to increase the conversions for online fashion stores.

As mentioned above, the team has worked extensively in the area of image recognition specifically for understanding fashion retail. The image recognition algorithm is being trained to judge colours, patterns and even style like the way a human does.

It leverages a hybrid methodology which uses AI to augment human capability and humans to augment AI where it is weak. “This multifaceted approach enables us to provide recommendations better than only humans or only AI.” says the team.

Product recommendations using image processing is a service being offered by a few other AI startups too, however the team isn’t dependent solely on this technology. “Streamoid uses a combination of a fashion rules set developed by leading stylists, machine learning, bid data analysis, image recognition, visual search and Stylist supervised machine learning to power it’s recommendation engines.” says Rajesh. The team at Streamoid is jointly using computer vision with machine learning, where the system is able to learn from the visitor’s preferences & also pick up trends from other buyer behaviour.

“We are able to return exact or similar matches for images under sub second from amongst very large datasets.” adds Rajesh.

You can discover a few case studies here.

A Virtual Stylist

Screen Shot 2016-11-24 at 2.54.01 PM

Its core technology has been deployed in two products –

Streamoid outfitter, which makes outfits based on occasions and Streamoid Stylebot which acts like a personal styling assistant.

The virtual stylist/shopping assistant sits on the retailer’s site and also works using NLP to help customers with their purchase, and further lets eCommerce retailers to cross-sell and increase the Average Order Value of each transaction.

In addition it can be also be accessed directly from several messaging platforms like FB messenger, LINE. Thus allowing retailers to increase the touch points.

The outfit recommendation engine has learnings from several in-house stylists and latest trends, so that it can recommend complete looks in real time.

Building a Business in Fashion AI

The company bootstrapped for a year while it was developing a prototype, and raised an angel investment of $1 million in late 2015. Its first customer went live in mid-2016 and currently has 6 clients in production and several in pipeline.

“At this point we are focused on growing business and rolling out new features to our Fashion platform. We are always searching for strategic investor partners who can help to grow the business using their industry expertise.” shares Rajesh.

The company is generating revenues, and targets to be profitable soon.

Challenges and Troubleshooting Them

When asked about the challenges they have come across, Rajesh says “Each day poses a different challenge for startups but finding good talent tops this list. The technology industry around us was using cheap investor money and spoiling the market by unreasonable salary hikes. Attracting quality talent who want to take the risk of working at a startup is a big challenge. The Indian startup risk appetite is still lacking but just beginning to grow!.”

The company uses all available resources for hiring. It presently has 17 full-time employees, which includes experienced people in mobile front-end and back-end technologies and AI/algorithms. In addition, they have experienced stylists/designers who have graduated from the top design/fashion schools.

Regarding other challenges, he adds “Closing enterprise sales is a big task because of the long lead times. Till we have a Live customers, we cannot refine the product. We have got live customers now and our products get refined every day.” A huge amount of data is required to enhance any AI & machine learning engine, especially if it has to match human capabilities.”.

About the Founders

Shown Above - Streamoid Founders, Sridhar Manthani (L) & Rajesh Kumar (R)

Shown Above – Streamoid Founders, Sridhar Manthani (L) & Rajesh Kumar (R)

Sridhar Manthani is the Chairman & CEO. He is a serial entrepreneur, mentor and investor. He started his career in Intel as an ASIC design engineer and worked on Microprocessor chips. He left Intel to be a part of the original team in a startup called S3. As Vice President of Engineering for S3 he grew his business unit to $500 million revenues. The Company went for a NASDAQ IPO in 1993. In 1997 Sridhar co-founded Thinkit Technologies Inc. Intel acquired the company in the year 1999. Post this he launched Nvidia India as their first employee in India and over eight years grew Nvidia India to 1000+ employees. He is an Angel investor in several startups based in India.

Rajesh Kumar is the CTO. He has been a part of great companies like Microsoft (MS accelerator (batch 01) ), Motorola (Core networks) ,Yahoo India R&D (core libraries) and InMobi. He has extensive experience in search and building big data platforms. He has worked on massive 4000 node (Hadoop) clusters to 8 bit (Ardiuno) micro-controllers; on core web server libraries to building video ad servers and search of all kinds. He has won many Hackathons.

Have ideas to share? Submit a post on iamwire

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>