Machine learning, LLMs, recommendations and applied AI.
How large language models help shoppers find what they mean, not just what they type. Author: Aarav Nigam Introduction Every search box eventually hits a wall where traditional search stops being useful. Searching tide powder 100g on Instamart works seamlessly. Exact keywords…
Contributor: Abhik Debnath In this article, we use Razorpay’s Return to Origin (RTO) risk model to explain the ML lifecycle: how we build a reproducible release, deploy it into checkout, monitor its behaviour, and use later outcomes to evaluate whether it needs retraining. We…
The model is only one part of the inference pipeline. At scale, the work happening around it, from routing requests to coordinating downstream dependencies, can have a much bigger impact on latency and efficiency. We addressed this by separating model execution from request…
Recurring subscription billing in India is deceptively complex. A single billing cycle has to send a mandated pre-debit notification, wait a regulator-defined window, then attempt collection by coordinating
India is undergoing a massive digital transformation, and at the heart of this revolution is PhonePe, enabling seamless payments and financial services for over 700 million registered users across Bharat. When you build at this kind of population scale, handling millions of…
3DS vs 3DS2 explained: risk-based authentication, frictionless flow, global mandates in the EU, India, Japan & Australia, and what changed after 3DS1's sunset
We run business-critical, long-running workflows (livestream lifecycle jobs, livestream user-experience jobs, recording and HLS streaming, etc.) on self-hosted, in-house Temporal clusters.
Global checkout design guide for lifting conversion with local payment methods, currency UX, smart routing, tokenization, card updater services, 3DS, and retries.
The Problem A Delivery Executive (DE) finishes their day, opens the app, and expects a clear breakdown of what they earned. Instead, they often see a payout that does not fully match their expectations — an incentive missing, a deduction they do not recognise, or a credited…
Network tokenization vs PCI tokenization: how they differ at the architecture level, what each does to authorization rates, fraud, PCI scope, and vendor lock-in
Part 1 of a multi-part series. A quick note before we begin: the first chapter of Khoj.AI was built before the generative AI wave really took off, so some early choices reflect the tooling of that time. This post walks through how the system has evolved over the last 1 year,…
Discover how Juspay built Genius, a reliable AI payment intelligence agent. Learn about agent harness, schema grounding, and multi-agent systems for accurate enterprise analytics.
Explores the shift to AI-driven Agentic Commerce and how Juspay powers secure, verifiable infrastructure for this intent-based evolution.
At ShareChat, we recently overhauled the architecture of our events ingestion system. As part of this exercise, we also migrated all our event data contracts from JSON to protobuf.
A payment vault is a digital database that stores sensitive information like credit card, bank account details, etc.
3DS 2.0 is updated to deal with current challenges, while enhancing security, improving user experience, and reducing friction during online transactions.
Explore biometric payment methods and their role in secure transactions. Learn about biometric payment systems, benefits, challenges, and the future of digital payments.
Google's Universal Commerce Protocol (UCP) enables AI agents to purchase across the web. Learn how payment orchestration solves multi-protocol commerce challenges.
Juspay’s SDKs are key tools that help merchant applications interact with our Orchestration Layer.
Juspay has built an internal AI-powered tool that eliminates guesswork, boosts efficiency, and aligns design with development.
The Auto-Transpiler is a semantics-preserving system that translates code between languages while maintaining consistent behavior across functions, execution paths, and database interactions.
At Juspay, we've distilled the complexity of agent-based systems into a simple, yet powerful showcase of how the agent loop fundamentally reimagines automation in payment processes.
We’ve built a suite of AI-assisted tools for designers to validate their work, for developers to meet visual standards, and for teams to measure design system adoption in real time.
Airborne is an over-the-air update platform which helps merchants eliminate release bottlenecks and prevent delays that could disrupt millions of transactions and users
Understand what a passkey is, how passkeys work, and why they’re replacing traditional passwords for secure, passwordless authentication. Learn more on Juspay's blog
At ShareChat, we recently overhauled the architecture for the data ingestion system. As part of this exercise, we also moved 100% of our event data contracts from JSON to Protobuf.
At ShareChat's Central Data Platform (CDP) team, we handle more than 1,000 different event types across 5 apps, ingesting upward of 80 billion events per day.
At ShareChat’s Central Data Platform (CDP) team, we cater to more than 1000 different event types from 5 different apps, receiving and ingesting more than 80 billion events per day on a regular basis.
At PhonePe Limited (“PhonePe”), the top banner on our platform’s homepage (the “masthead”) isn’t just an ad slot—it’s the important real estate we have. However, delivering real-time, personalized recommendations at a scale of hundreds of millions of users presents a complex…
At ShareChat, Redis APIs were embedded in almost everything: caches,recommendation delivery systems,deduplication queues, ranking pipelines, counters, you name it.
By mid-2023, we had transformed our ML stack—building a real-time feature store, optimizing model retrieval, and fine-tuning ranking. But two critical gaps remained: 🔹 Scaling model inference without hitting infrastructure roadblocks 🔹 Moving embedding search from batch to…
At ShareChat — a leading social media platform with over 200M monthly active users (MAU) — our data platform ingests and processes more than 1 million events per second. However, this event load isn't constant.
By late 2022, we had built something we were truly proud of—a real-time ML serving system with a DAG-based executor, a feature store, and an interaction store powering key ranking and personalization models. It was a major milestone, the culmination of months of effort from data…
Every time a user views a post, an event is fired to inform us. We feed this real‑time stream into our recommendation system, e.g. analytics pipelines, feature computations, and view counters.
Search at Meesho For an e-commerce platform like Meesho, optimizing Search is pivotal to enhancing the user experience, improving product visibility and faster discovery, and eventually increasing conversion rates. Users search in multiple Indian languages looking for quality…
At ShareChat, our in-house ML feature system powers personalized feed recommendations across the platform.Built on Apache Flink, it serves as a feature-as-a-service platform for scalable computation
The Genesis: How a Friday Night Roast Sparked Meesho’s ML Platform It all started in early 2022, over a casual Friday evening catch-up. Like many great origin stories, this one began with friendly banter between a group of backend engineers and data scientists. As the…
Revolutionizing Communication at Scale At Meesho, seamless communication is at the heart of our user experience. Whether it’s an OTP for login, an order update, or a personalized shopping recommendation, our platform ensures timely and reliable messaging across multiple…
Revolutionizing Communication at Scale At Meesho, seamless communication is at the heart of our user experience. Whether it’s an OTP for login, an order update, or a personalized shopping recommendation, our platform ensures timely and reliable messaging across multiple…
Our recording infrastructure focuses on WebRTC video and audio. The goal is to allow moderation teams to access clean, archived footage—composite or individual—for compliance, post-production, or analysis.
At ShareChat, Machine Learning models form the backbone of our recommendation system, playing a pivotal role in delivering the most relevant and engaging content to our users.
The primary goal of a recommendation system is to identify and deliver the most relevant content for a user. To achieve this, it must not only identify high-quality content but also ensure diversity
Modern recommender system models can be very compute- and memory-intensive and expensive to serve. In this post, we’ll go over some of the things we did at ShareChat to make our inference more cost effective.
In recent years, online shopping has surged, revolutionizing how people purchase products and services. E-commerce’s convenience has reshaped consumer behaviour and the retail landscape. Unlike traditional stores, online shoppers often face sizing challenges, leading to…
At ShareChat, delivering personalized recommendations to millions of users requires building and training large-scale recommendation models.
Sharechat is India’s largest homegrown social media company with native and regional user generated image and video content. Our two apps, ShareChat and Moj, are used by over 325 million users in 15 regional languages across India.
Over the past few weeks, we’ve introduced a series of exciting updates to PromptQL that enhance its power, accessibility, and collaboration capabilities. These enhancements are focused on simplifying workflows. Everything is now live and ready to explore!
ShareChat's Livestream redefines creator-audience bonds, blending entertainment, monetization, and genuine connections in a dynamic social media ecosystem.