Khalti Digital Wallet
Khalti
1.5M
Daily transactions managed
8-10M
Registered users on the platform
$10M+
Daily transaction value processed
60%
Reduction in fraud losses
Overview
Khalti is one of Nepal's leading digital wallets, serving an estimated 8 to 10 million registered users in a country of 30 million people. As Lead Product Manager, I owned the product strategy for both the consumer wallet and merchant platforms, covering wallet funding, bill pay, peer-to-peer transfers, merchant checkout, refunds, and payouts for roughly 1.5 million daily transactions.
I joined during what was arguably the most consequential growth phase in Khalti's history. The COVID-19 pandemic had accelerated digital adoption across Nepal, and the habits formed during 2020 lockdowns were being cemented into permanent behavior. My mandate was to evolve the platform from rules-only flows to AI-powered personalization and risk scoring, a transition that mirrors what platforms like M-Pesa, bKash, and GCash underwent during their own scaling phases.
Nepal's digital payments ecosystem was growing at 50 to 100 percent year-over-year during this period, with the total value of digital transactions growing from approximately NPR 3-4 trillion to over NPR 10-12 trillion. Operating in this environment meant balancing aggressive growth targets with increasingly complex regulatory requirements from Nepal Rastra Bank, the country's central bank.
The Problem
At scale, rules-based systems for fraud detection, personalization, and risk scoring break down. Fraud patterns evolve faster than manual rule updates. Personalization at scale cannot be hand-coded for millions of user segments. Transaction approval and decline decisions need to balance fraud prevention with user experience, which requires ML models that can learn from patterns. Khalti needed to transform its infrastructure from deterministic rule-based systems to probabilistic AI-driven systems while continuing to serve millions of users daily without disruption.
My Role
Lead Product Manager
I owned the end-to-end product strategy for consumer and merchant platforms. This included leading discovery and rollout of recommendation engines for services, billers, and offers based on behavioral and transaction data. I partnered with data science, risk, and compliance teams to design onboarding and KYC flows using machine learning, and worked with engineering and bank partners to embed real-time fraud and credit risk models into payment APIs.
The Approach
I defined experimentation frameworks, KPIs, and dashboards for the payments funnel, tracking payment success rate, decline reasons, fraud rate, chargeback ratio, and cohort behavior. A/B tests on pricing, rewards, and in-app journeys helped optimize outcomes while maintaining compliance.
The transition to AI-powered systems was phased: first, we instrumented the platform to capture behavioral and transaction data suitable for ML training. Then we piloted models alongside existing rules in shadow mode. Finally, we gradually shifted decision-making to ML models with human oversight for edge cases. Throughout, I coordinated post-launch monitoring, setting thresholds and alerts for spikes in declines, fraud, or latency.
Collaboration with UX, customer experience, and engineering teams improved user and merchant experiences for declined transactions, account limits, and risk-based holds. Clear communication about KYC checks, limits, and refund timelines reduced support tickets and built trust.
Key Features
What we built
AI-Powered Personalization
Recommendation engines for services, billers, and offers based on behavioral and transaction data, improving payment conversion and merchant revenue.
Real-Time Fraud Detection
Machine learning models embedded into payment APIs for real-time transaction scoring, replacing static rule-based systems.
Smart KYC Flows
ML-driven onboarding that flags high-risk applications, predicts drop-off, and dynamically adjusts friction based on risk profile.
Payments Analytics Dashboard
Comprehensive funnel tracking with payment success rate, decline analysis, fraud rate, chargeback ratio, and cohort behavior visualization.
Merchant Platform
Full merchant ecosystem including checkout, settlements, refunds, and payouts with real-time transaction monitoring.
A/B Experimentation Framework
Systematic experimentation on pricing, rewards, and in-app journeys with statistically rigorous measurement of outcomes.
Tech Stack
Key Lessons
What I took away from this project
Incremental migration from rules to ML beats big-bang rewrites every time
Invest heavily in observability before scaling — you cannot fix what you cannot see
Customer communication during incidents builds more trust than perfect uptime
Technical debt in payment systems compounds faster than in most other domains
More Projects
Explore other work
Want to build something similar?
I help companies scale their products and build high-performing teams. Let's discuss how I can help with your next project.
Get in Touch