India’s leading fashion e-commerce player Myntra is leveraging artificial intelligence (AI) and machine learning (ML) solutions to level up the overall customer experience.
The Walmart-owned company has recently launched a short video platform ‘Myntra Minis’ to improve user engagement by offering engaging visuals and dynamic content. This specific feature enables customers to enjoy phygital presentation of every kind of product available on the platform.
Moreover, the firm has integrated Chat GPT (Generative Pre-trained Transformer) into its search capabilities to optimise product listings for increased visibility on the platform.
Myntra engages 15 million users with products from over 1000 brands through its social commerce platforms. It aims to deliver a “superior” customer experience from browsing to post-purchase support.
The company said it uses in-house AI technology, ML models and computer vision algorithms to power many new features like My Stylist, Fashion Object Detection, Image Search and Outfit Recommendations.
For instance, the AI-based My Stylist feature personalises the home screen, and recommends outfits, sizes and designs in real-time for customers based on their previous purchases.
These aspects use an AI algorithm that taps into the platform's catalogue data via deep convolutional neural networks and is also trained using a Bi-directional long short-term memory (Bi-LTSM) model that expands dataset from Myntra's catalogue, which enables the feature to handle a high volume of requests.
Data architecture
While traditional data platforms often struggle to support modern applications, Myntra’s senior architect Narayana Pattipati told the Aerospike Summit 2023 that his company has deployed a modern data platform with unique Hybrid Memory Architecture (HMA) powered by Aerospike that combines flash storage and RAM to provide speed, maintain scalability and cost-effectiveness of disk-based storage.
With Aerospike, Myntra has been able to consolidate diverse data streams into one database for the entire ML lifecycle and also scale issues whenever new ML use cases are onboarded. It also supports our performance and scale requirements, Pattipati said.
Aerospike optimises online feature stores, state stores and inference cache for Myntra to solve many business problems.
For instance, Myntra optimises online feature stores to solve issues in widget ranking, which helps to improve customer experience and engagement, thus increasing revenues.
Myntra plans to invest further in data science, enabling the majority of its new features to go through AB testing, a methodology for comparing two versions of a web page or app against each other to determine which one performs better, Pattipati said.
“This helps us create a significant impact on customer experience, providing real-time and offline solutions with varying latency requirements,” he added.
ChatGPT integration
Myntra's new product discovery feature - MyFashionGPT is powered by OpenAI’s large language model ChatGPT 3.5. It helps users make natural language-based queries for product discovery.
The queries can be related to any occasion, popular celebrity looks, global or local events, and destinations.
Developed in-house, this feature processes responses from Myntra’s search ecosystem to show curated lists of products across categories to the user.
The AI assistant can also refine its results based on follow-up queries, based on internet-trained datasets and on semantic understanding of the user's query, Myntra said.
In the near future, Myntra plans to work on voice search and conversational interaction features to offer more personalisation.
