AI Agents Take Autonomous Actions: Leveraging Apps through Marketplace MCP Gateway
By Kamal Kumar, Rahul Ganesh Partheeban To enable AI agents to autonomously execute real-time business actions across third-party applications, we built and productionized the Model Context Protocol (MCP) Gateway at freshworks. This gateway allows AI agents to securely discover…
From Rigid Rules to Empathetic Agents: Rebuilding Swiggy’s Order Cancellation Experience with LLMs
IWTCIntroduction If you have ever typed “I want to cancel my order” into a support chat, you were probably not looking for a workflow. You were looking for clarity, speed, and some assurance that the platform understood what had gone wrong. At Swiggy, this seemingly simple…
Explaining Every Rupee: How We Built Reliable LLM Support Bots for Delivery Partners
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…
What Happens When You Turn 150+ Icons Into One Font
by Madhankumar J The cost of “simple” icons This is part of the Unity Craft series, where we share how we’re improving our application — performance, tooling, and consistency. In this instalment we talk about icons. Icons are everywhere in modern UIs — buttons, navigation,…
How Freshworks is building software without writing software by Ajay Ravichandran The software industry is approaching an inflection point. For decades, the measure of engineering productivity has been the same — how fast can a team ship well-tested, well-reviewed code? The…
The System Behind Every Seller Payout: How Meesho Rebuilt Trust at Scale
Every morning, lakhs of Meesho sellers open their payment tab looking for one simple thing: confirmation that the orders they fulfilled have translated into money they can trust. For a seller, a payout is not just a bank transfer. It is a working capital. It is confidence to…
Teaching a Geocoder What Every Delivery Agent Already Knows
This is Part 2 of our GeoIndia series, which follows Part 1’s introduction of GeoIndia, the Seq2Seq model that converts messy Indian addresses into precise map locations. In this post, we go one level deeper: what we learned from running GeoIndia in the wild, and how that pushed…
Built by Many: How ScriptedByHer Is Reshaping the Face of Tech at Meesho
The concept of hackathons is well-established and has proven to be very popular among tech enthusiasts across India. These intensive two or three-day events bring people together who are passionate about technology to work on software-related challenges, often provided by…
When the Wrong Thing Comes Back: How Meesho Stopped Putting the Burden on Sellers
Imagine you're a seller on Meesho. You shipped a bedsheet. A few days later, a return arrives at your door. You open it and find a bar of soap. What comes back isn't what was shipped. The seller is left with dead inventory, and to be compensated, they have to prove what went…
Network Tokenization vs PCI Tokenization: Differences, Architecture, and Which One Lifts Authorization Rates
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
The Quiet Overhaul: How Meesho Stopped Answering the Same HR Questions Twice
There's a moment every HR professional knows well. It's 2:30 on a Monday, you're mid- way through something that actually matters, and someone pings you asking what the casual leave policy is. You answer. An hour later, someone else asks the same thing. This isn't a failure of…
When processing billions of transactions in a system as large as PhonePe, underlying databases go through constant querying, repairs and management. While a set of these tasks are automated, some are managed by human users. This leaves a potential risk of unauthorised/untracked…
Finding the Needle in the Stack: How We Built Khoj.AI for Automated Root Cause Detection
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,…
Building a payment intelligence agent enterprises can actually trust
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.
Building a Modern Performance Testing Suite with Gatling, Gradle, and CI
by Ravi Prasanth At Freshworks, performance is a critical launch requirement, not a secondary concern, especially for products with heavy reliance on analytics and reporting workflows used by customers at scale. As the scale of our reporting and user concurrency increased, our…
The Signal in the Noise: Scaling Observability with SLO-Driven Alerting
Author: Kartik Ukhalkar In a typical SOA, if your service is misbehaving, what is the most efficient way to detect it while minimising the noise? Whether you are an administrator or developer, you would have thought about this problem a few times. We asked ourselves the same…
Beyond Imitation: Teaching AI to Speak Like a Character in Hindi
Introduction As JioHotstar serves a vast, diverse audience, generative AI is crucial for fast, scalable content creation across languages and formats, capabilities that traditional methods cannot match. Content systems are rapidly shifting to generative-first architectures,…
Teaching a Machine to Read Indian Addresses: How We Built GeoIndia
GeoIndia: A Seq2Seq Geocoding Approach for Indian Addresses Download to access the research paper GEOINDIA_V1.pdf 5 MB download-circle Type an address into any major mapping service and you'll get a pin dropped on a map in seconds. It feels effortless. But under the…
The Pulse of India’s Supply Chain: How Flipkart Plans for 500 Million Customers — Part 1
In the world of e-commerce, a click of a “Buy Now” button starts an invisible race. Across India (metro cities or remote villages) millions of customers expect their packages to arrive with a speed that feels like magic. But behind that “magic” is a monumental logistical puzzle.…
Ordering Biryani Without Missing a Six: The Swiggy × JioHotstar Story
Picture this: Your favorite batsman is on strike, two runs needed off the last ball — and you’re starving. Continue reading on Swiggy Bytes — Tech Blog »
When Graphs Aren't Enough: How We Used Hypergraph Neural Networks to Predict Return-to-Origin at Meesho
A Dual Perspective Hypergraph Neural Networks Download to access the research paper DPHGNN.pdf 2 MB download-circle Every e-commerce company has a version of this problem: an order gets placed, the delivery is attempted, and the package comes back. Return-to-Origin, or…
Razorpay Oncall Agent: From 30-Minute Investigations to 90-Second AI Analysis
Our on-call engineers were spending 30 minutes investigating every production alert. Here’s what happened when we automated it. At 3 AM, alerts don’t care about your sleep schedule. When our payment infrastructure threw an error last month, our on-call engineer spent 32 minutes…
Migrating 1000+ event contracts from JSON to Protobuf
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.