Cluster Manager - Digital Platforms and Consumer AI/Senior Cluster Manager - Digital Platforms and Consumer AI

Location Name: Pune Corporate Office - Mantri

Job Purpose

At Bajaj Finance, we help millions of people across India manage their money. People come to us when they want to buy a new phone on EMI, when they need a personal loan for a wedding, when they want to buy insurance, or when they want to grow their savings in a Fixed Deposit. Because we have so many customers, we get thousands of questions every single day. Customers want to know their account balance. They want to change the date their EMI gets deducted. They want to know the fees for closing a loan early.

In the past, to get these answers, a customer had to wait on hold to talk to a call center agent, or they had to tap through ten different screens on our mobile app. We want to change that. We want to make it as easy as sending a text message to a friend.

We are building a massive, highly intelligent AI (Artificial Intelligence) platform. Think of it like a super-smart chat assistant built right into the Bajaj Finance app. We want our customers to be able to just type out their problems in plain English, Hindi, or other languages, and have the AI fix the problem instantly. If a customer types, "Hey, I need to pause my loan payment this month," the AI should understand them, check their account, and do the work in seconds.

Building this is incredibly hard. It is not just a simple robot that gives canned answers. It is a smart system that has to understand human language, connect to our deep banking systems, and follow strict financial rules.

Duties And Responsibilities

Your Daily Work: The 9 Core Areas

About

Your job will be very hands-on. You will not just be sitting in long meetings talking about big ideas. You will be doing the real, detailed work to make this AI smart. Here are the nine specific areas you will be working on every day:

• FPR (Future Press Release)

Before a single line of computer code is written, you have to know what your end goal is. We do this by writing a Future Press Release, or FPR.

A press release is a news article that a company puts out when they launch a cool new product. But here, you will write this news article before we even start building. You will write it as if the product is already finished and launched to the public. You will write a catchy headline like: "Bajaj Finance Launches AI That Solves Loan Issues in 10 Seconds." You will write quotes from pretend customers talking about how much they love the feature. You will explain exactly what problem the feature solves in simple, everyday language.

Why do we do this? Because it keeps everyone focused. If the news article sounds boring or confusing, it means your idea is probably boring or confusing. Once you write a great FPR, you show it to the engineers and your bosses. When everyone reads it and says, "Wow, this sounds amazing, let's build it!"—that document becomes your team's guide.

• FAQ (Frequently Asked Questions)

Along with your news article, you will write a massive list of Frequently Asked Questions (FAQ). But these are not just the normal questions a customer might ask. These are questions that anyone involved might ask.

You have to put yourself in the shoes of a confused customer. What will they ask? "Is my bank data safe with this AI robot?" You write the answer. You have to think like our call center agents. "Will this AI do my job wrong and make the customer angry?" You write the answer. You have to think like our company lawyers. "Does this AI follow the Reserve Bank of India rules?" You write the answer.

The goal of this FAQ is to force you to find all the problems and risks before the engineers start building. If you cannot answer a tough question in your FAQ document, it means your plan has a hole in it, and you need to fix your plan.

• PPG (Product Program Guidelines)

Because Bajaj Finance deals with people's money, we have a lot of strict rules. We cannot just give a loan to anyone. We have to charge exact fees. We have to use exact interest rates. All of these strict rules are written down in company rulebooks called Product Program Guidelines, or PPG.

Every single product has a PPG. The personal loan has a PPG. The fixed deposit has a PPG. When we build the AI, the AI must follow the rules in the PPG perfectly. It cannot make a mistake. If the PPG says the late fee is 500 rupees, the AI cannot accidentally tell the customer the late fee is 400 rupees. That would cause a massive legal problem.|As a Product Manager, you will read these PPG rulebooks. You will pull out all the rules related to your project. You will work with the engineers to make sure these rules are permanently locked into the AI's brain. If the company changes a rule in the PPG next year, you are the person who makes sure the AI gets updated immediately.

• Synthetic Testing

Once the engineers build a version of the AI, we have to test it. But we cannot let a broken, untested AI talk to real Bajaj Finance customers. That is too risky. If the AI is confused, a real customer might get bad advice about their loan.

So, we use something called Synthetic Testing. "Synthetic" just means fake or made-up. You will help create thousands of fake conversations to test the AI safely. You will create fake customer profiles. For example, you make a fake customer named Rahul who has a home loan. You will type to the AI pretending to be Rahul. You will ask, "Can I pay off my whole loan today?" and see what the AI says.

Does the AI say yes? Does it calculate the early payment fee correctly? What if you pretend to be a customer who types with terrible spelling? What if you type in a mix of Hindi and English? What if you pretend to be very angry? Does the AI still understand and stay polite? You will run hundreds of these fake tests. When the AI fails, you write down why it failed, take that information back to the engineers, and tell them what to fix. You keep doing this until the AI passes every fake test perfectly.

• LLM Parameters (What We Send to the AI Brain)

The core brain of our AI is called a Large Language Model (LLM). Think of the LLM like an incredibly smart person sitting in a dark room. This person knows how to speak perfectly, but they have no idea who they are talking to.

If a customer types, "What is my balance?", the smart person in the dark room has no idea whose balance to check. So, we have to send the LLM a package of extra information along with the customer's question. We call this setting the LLM Parameters.

As the Product Manager, you will decide exactly what information goes into that package. When the customer asks about their balance, you will tell our computer systems: "Before you let the AI answer, grab the customer's name, grab their last three payments, and grab their current loan amount. Send all of that to the AI."

You also have to send strict instructions. You will tell the AI, "You are a helpful assistant for Bajaj Finance. Always be polite. Do not guess numbers. Only use the numbers I just sent you." If you send the AI too much information, it gets confused and it costs the company too much money to process. If you send too little information, the AI gives a stupid answer. Your job is to find the perfect balance of what data to send the AI every time a customer speaks.

• Visualisations

Chatting with an AI should not just be reading massive blocks of text. Nobody wants to read a giant paragraph of words on their small phone screen.|Sometimes, the best way to answer a customer is by showing them a picture, a chart, or a button. We call these Visualisations.

Let's say a customer asks, "How much of my loan have I paid off?" The AI could write a long, boring sentence with a bunch of numbers. But it is much better if the AI shows a nice, clean pie chart on the screen. Half the pie is green for the money paid, and half is grey for the money left. The customer understands it instantly.

Or, if the AI needs the customer to pick a date, it shouldn't make the customer type "October 15th." It should pop up a small calendar on the screen so the customer can just tap the day. You will decide when to use text and when to use these visual tools. You will work with our designers to sketch out how these charts and buttons should look so the chat is beautiful and easy to use.

• Product Switching

Imagine you are talking to a human salesman at a store. You ask about buying a TV. Then, suddenly, you change your mind and ask about buying a fridge. The human easily switches the topic to the fridge.Computers are not naturally good at this. If the computer is programmed to talk to you about a TV, and you ask about a fridge, the computer will get totally confused. It might say, "Error, I do not understand."

At Bajaj Finance, we have dozens of products. A customer might start a chat asking about their EMI shopping card. But halfway through the chat, they might say, "By the way, I also want to open a Fixed Deposit." The AI needs to smoothly switch from talking about the shopping card to talking about the Fixed Deposit without breaking or crashing.

We call this Product Switching. You will write the rules for how the AI handles these sudden changes in conversation. You will plan out how the AI remembers what the customer was talking about first, just in case the customer wants to go back to the original topic later.|8. Conversation Flows

A normal conversation flows naturally. But computers need a map. They need to be told exactly how a conversation should go, step by step. You will create these maps, which we call Conversation Flows.

You will literally use a software tool to draw boxes and arrows on your screen.

• Box 1: The AI says, "Hello, how can I help?" -> Arrow points to Box 2.
• Box 2: The customer says, "I want to see my statement." -> Arrow points to Box 3.
• Box 3: The system checks if the customer's phone number is verified.

o If Yes -> Arrow points to Box 4.

o If No -> Arrow points to Box 5.

• Box 4: Show the statement on the screen.
• Box 5: Ask the customer to type in a one-time password (OTP).

You have to map out every single path the conversation could possibly take. You have to think of all the "what ifs." What if the customer types the wrong password? What if the customer stops replying and walks away from their phone? What if the internet crashes? You must draw a box and an arrow for every single situation. These flowcharts become the actual instruction manual for the engineers. They look at your map and write the code to make it happen exactly as you drew it.

Required Qualifications And Experience

The Background You Need (Education & Experience)

Requirements

Because this job requires you to build highly complex computer systems, we are very strict about the background you need to have. We are looking for people who meet these exact requirements:

• Your Education (The Degree and the College)

You must have a Bachelor’s degree in Computer Science, Software Engineering, or a closely related technical computer field.

You must have graduated from a Tier 1 Engineering College. This means premier institutes like the IITs (Indian Institutes of Technology), NITs (National Institutes of Technology), or top-level Regional Engineering Colleges.

It is because the AI technology we are building is extremely complicated. You will be working side-by-side with some of the smartest software engineers in the country. You need to know how software is actually built. You need to understand coding logic, databases, and how computer servers talk to each other. Top engineering colleges train you to handle these tough, highly technical problems. If you do not have this strong computer science foundation, you will not be able to keep up with the engineering teams.

If you also have a Master's degree, that is great! We love to see it. But if you do not have a Master's degree, that is perfectly fine. It is not required. The Bachelor's degree from a Tier 1 college is what matters the most.

• Your Work Experience (2 to 3 Years in NLP/AI)

You cannot be fresh out of college for this job. You must have at least 2 to 3 years of actual, hands-on work experience.

More importantly, this experience must be in a very specific area. You must have worked on NLP platforms. It is the specific technology that teaches computers how to read, understand, and write human language.

You must have spent the last few years of your career actually building chatbots, voice assistants, or smart AI text tools. You cannot just be someone who managed a normal website or a standard mobile app.