LEAD PRODUCT MANAGER · BENGALURU
Products built for millions.
Decisions grounded in outcomes.
I’m Mahendra, a Lead AI Product Manager with 8.9 years of experience building and scaling conversational AI, search, CX automation and ONDC commerce products. I’ve led product strategy for RAG-powered chat and semantic search, partnering with engineering on prompts, evaluation and guardrails, and using A/B testing to drive measurable outcomes.
Figures are project outcomes from the periods noted in the case studies.
SELECTED WORK / 2017—PRESENT
Five products. Different problems.
One focus on measurable change.
Scan the outcomes, then open the work that matters most to your team. Detailed historical material remains available in the original portfolio PDF.
Turning support chat into a service and revenue channel
GPT-4o, RAG and semantic retrieval across 14 financial verticals.
Apr–Dec 2025
Making financial products easier to find
Semantic ranking and contextual discovery across app and web.
Growing an ONDC-native buyer experience
Repeat purchase, checkout and marketplace performance.
Resolving payment issues at scale
Guided, multilingual support in the transaction journey.
Giving support agents better tools
Transaction-aware workflows and one-click resolution.
CONVERSATIONAL AI · RAG · FINTECH
From service bot to product channel.
The challenge was to serve and sell across many financial journeys without losing answer quality. I led product strategy and requirements for a GPT-4o and RAG-powered chat experience integrated with Azure Cognitive Search across 14 verticals.
Product decisions
- Evolved retrieval from GPT-3.5 hierarchical routing toward Azure Cognitive Search-based semantic retrieval.
- Co-developed prompt framing, edge-case coverage, output constraints and numerical guardrails for financial answers.
- Balanced service automation with sales journeys and human escalation.
What changed
- 81% fewer agent-handled requests with 96.2% bot accuracy.
- 405% year-over-year chat adoption growth; 5.1-second response latency while scaling.
- ₹38.3 Cr in loan disbursements and about 25,500 EMI card disbursements during Apr–Dec 2025.
The original portfolio contains the earlier product walkthrough, features and screens. Its historical metrics describe an earlier stage of the product.
From intent to an answer a customer can act on
These supplied screens show the assistant recognising the customer’s card or loan relationship, offering relevant topics, answering an EMI question and guiding a follow-up withdrawal task. The UI details remain exactly as captured.
SEARCH & DISCOVERY
Helping customers find the right next step.
Search spans many financial products and intents. I owned the horizontal Search product across key app and web pages serving 45M+ monthly active users.
Product decisions
- Prioritized semantic ranking, personalization and contextual discovery.
- Introduced Trending Searches and Recent Searches to reduce zero-query friction.
- Used structured A/B testing to validate changes.
What changed
The measured search exit rate fell from 24.7% to 9.32%. The experience supported discovery across a platform with 45M+ monthly active users.
From an open search to a useful next step
These four supplied mobile screens show different search intents. Recent and trending terms help people start; service queries route to relevant actions; product queries lead to browsable results. The captures illustrate the journey, while the exit-rate result above is a separate measured outcome.
One search entry across surfaces
The public desktop header shows where the same broad discovery problem begins on web. Its current interface may differ from the product period described above.
ONDC · E-COMMERCE
Building for repeat purchase on an open network.
Since January 2026, I’ve owned Hamara Mall, Bajaj Finserv’s ONDC-native commerce product, with responsibility for product priorities across grocery, electronics, fashion and other categories.
Product decisions
- Shipped Order Again, Multi-Seller Cart & Coupon, Source-Based Banners and Default Address Selection.
- Sequenced GMV growth, seller onboarding and order quality initiatives with RICE-based prioritization.
- Worked across engineering, catalog, logistics and compliance.
What changed
The platform contributed 11.3% of India’s ONDC retail orders in FY26. Search latency and core web vitals improved by 62%, supporting lower bounce and better discoverability.
Discover, browse and act on an offer
The homepage screens show the category breadth and merchandising surfaces. The Deals screens show the promotion-to-product path, with product cards and add actions visible without switching to a desktop view. Prices and offers are specific to the capture date.
Supplied mobile captures from hamaramall.com. Screens are presented without changing their content.
PAYMENTS · CUSTOMER EXPERIENCE
Support that starts in the transaction.
Customers needed help with payment outcomes without repeating transaction details to an agent. I designed guided, transaction-aware support flows across consumer, merchant and agent channels.
Product decisions
- Placed support in the transaction screen with status-aware summaries, guided issue choices and escalation paths.
- Introduced a 48-hour in-app status check for pending UPI transactions.
- Used multilingual journeys to improve accessibility for consumers, merchants and agents.
What changed
- Automation improved from 78% to 95%.
- Pending-transaction tickets fell by 30%; agent fallback scenarios fell by 15%.
- Customer satisfaction reached approximately 72%.
The original portfolio contains the full flow and product screens.
Help grounded in the payment context
The supplied screens show a support journey for a wrong-account payment: a context-aware summary, guided issue selection, and a resolution response with bank follow-up links. Their on-screen date is 2024; the product work described above covers 2020–2023. The original capture content is unchanged.
AGENT EXPERIENCE · OPERATIONS
Make the right resolution easier to deliver.
Agents needed to interpret transaction context, choose a response and update tickets quickly and consistently. I designed the Unified Agent Panel around guided workflows and transaction-aware support.
Product decisions
- Connected ticket context and transaction details to guided issue selection.
- Used curated responses, one-click actions and automatic ticket-property updates.
- Added an agent feedback loop for response quality and workflow improvement.
What changed
The panel reduced average handle time by 35% and saved 91 man-days per month. Agents spent less effort on repetitive steps and more on complex customer issues.
The original portfolio preserves the detailed workflow and screens.
A guided path through the agent’s work
The Unified Agent Panel is a desktop Freshdesk experience. These three screens show the progression from issue classification to transaction-aware choices and a one-click response with ticket-property updates.
HOW I WORK
Strategy, delivery and the details between.
Define the customer friction
Start with the user journey, operational constraints and the metric that would prove improvement.
Make trade-offs explicit
Use evidence and prioritization frameworks to sequence work across engineering, design and business partners.
Measure after launch
Track adoption, quality, revenue and cost outcomes, then iterate on what the data shows.
ABOUT
Product leadership across AI, fintech and commerce.
Currently Lead Product Manager at Bajaj Finserv in Bengaluru. Previously Senior Product Specialist at PhonePe and Automation Specialist / Product Analyst at Virtusa for CitiBank. I manage eight APMs and work across product strategy, roadmaps, experimentation and cross-functional delivery.
The PDF is a historical archive with earlier-stage metrics and a 7.3-year biography. This site uses the current confirmed 8.9-year profile.