HomeAboutExperiencePortfolioToolsResearchBlogCoursesNewsletterServicesAwardsShopContact
Theme

Career Journey

Professional Experience

From building Nepal's largest digital wallet to architecting ML model serving and LLM/RAG systems at Morgan Stanley, and now shipping adaptive learning products at Tilli — my work is about building across ML, engineering, and product.

CurrentRemote

Product Builder

Tilli (TilliKids)

AI-powered adaptive learning platform

December 2025 - Present

  • Architect and build an AI-powered adaptive assessment platform end-to-end — shaping the product surface, writing the services behind it, and shipping ML inference on AWS Lambda/EC2 with pytest, Docker, and GitHub Actions CI/CD
  • Design high-throughput REST APIs and event-driven microservices for real-time scoring and content delivery using Python, FastAPI, PostgreSQL, Redis, and Celery, with telemetry via Amplitude and Mixpanel
  • Implement adaptive learning algorithms using gradient-boosted models and contextual bandits for personalized content sequencing, evaluating through offline A/B testing harnesses and learning-efficacy metrics
  • Own product decisions alongside engineering work — translating learner outcomes into model objectives, API contracts, and experiment design instead of handing specs over a wall

Technologies

PythonFastAPIPostgreSQLRedisCeleryAWS LambdaDockerML/AIA/B Testing
PreviousSan Francisco, CA

Product Builder

Morgan Stanley

Global Investment Bank

April 2024 - November 2025

  • Architected end-to-end ML model serving infrastructure in Python and Go for an AI risk analytics platform supporting 500+ institutional clients and $2T+ AUM — microservices for feature engineering, distributed training, batch and real-time inference, and automated drift detection
  • Designed REST and gRPC API layer for model serving with 99.9% uptime SLA, engineering load balancers, circuit breakers, retry policies, and observability via Datadog, Grafana, and Prometheus
  • Built LLM-based agent pipelines using RAG architecture with pgvector, LangChain, and the OpenAI API — engineering retrieval, re-ranking, and generation stages that reduced report generation latency by 75% under strict access controls
  • Owned the roadmap alongside the build — writing user stories with latency, explainability, and observability requirements, and partnering with InfoSec and compliance to meet firm policy and regulatory expectations
  • Automated ML pipeline orchestration reducing deployment cycles by 60% through GitHub Actions CI/CD, containerized training on Docker and Kubernetes, feature store integration, and model registry with versioning and rollback
  • Implemented model calibration and continuous retraining systems reducing false positive rates by 35% across 12 asset classes via threshold optimization, statistical evaluation harnesses, and automated A/B evaluation frameworks
  • Shipped Python SDK, API playgrounds, and integration test suites for 15+ client teams — accelerating onboarding by 40% via self-service sandboxes, contract testing, and reference notebooks

Technologies

PythonGoREST APIsgRPCLLMsRAGpgvectorLangChainKubernetesDatadog
PreviousLalitpur, Nepal

Product Leader

Khalti

Top Digital Wallet of Nepal

May 2021 - August 2023

  • Led the consumer wallet and merchant platforms for Nepal's leading digital wallet (3M+ MAU, $10M+ daily volume), owning product strategy and writing the backend services alongside engineering — P2P, bill pay, merchant checkout, refunds, and payouts for ~1.5 million daily transactions in Python, PostgreSQL, Redis, and Kafka
  • Designed a payment routing engine across 8 banking partner APIs, improving transaction success rate from 89% to 97% through intelligent retry logic, failover cascades, connection pooling, and real-time health monitoring
  • Built a real-time ML fraud scoring pipeline embedded in payment APIs using scikit-learn, feature engineering pipelines, and event-driven architecture on Kafka — reducing fraud losses by 40% and chargebacks by 55% via dynamic threshold tuning
  • Developed a recommendation engine using collaborative filtering and behavioral event streams, driving 85% YoY user growth and 120% transaction volume increase; built experimentation infrastructure on LaunchDarkly enabling 50+ annual A/B tests
  • Shipped ML-powered KYC/AML verification services with risk scoring models, reducing manual review by 60% for 500K+ users annually; built real-time observability dashboards in Tableau and Power BI tracking fraud rate, decline codes, and settlement metrics
  • Defined experimentation frameworks, KPIs, and incident reviews — tying payment success rate, decline reasons, fraud rate, and chargeback ratio directly back to model and product decisions

Technologies

PythonPostgreSQLRedisKafkascikit-learnPayment APIsA/B TestingKYC/AMLPCI DSS
PreviousKathmandu, Nepal

Product Builder

Tekkon Technologies

Australian & Global Product Services

April 2019 - May 2021

  • Led product and early engineering for Manini Pay — a cloud-native digital wallet and payments platform for merchants and consumers — from discovery through launch
  • Defined data collection and event schemas for ML pipelines powering fraud detection, risk scoring, and customer lifetime value, and partnered with data scientists to pilot merchant churn and high-value-merchant models
  • Exposed payment status and risk scores through REST APIs and webhooks for real-time partner dashboards, building the integration surface alongside the product spec
  • Shipped developer-friendly documentation, Postman collections, and sample apps in React and Python — treating developer experience as a first-class product surface

Technologies

ReactPythonREST APIsWebhooksMLPostgreSQL
PreviousLalitpur, Nepal

Product Builder

Smartpalika Co

Government Tech Software Provider

September 2018 - March 2019

  • Shipped e-governance modules end-to-end — citizen registry, online tax payments, and permits — translating policy requirements into product specs and working side-by-side with engineering on secure, localized payment flows
  • Facilitated workshops with municipal leaders and finance teams, turning fieldwork into backlog priorities and sprint goals
  • Supported deployments across 20+ municipalities and ran training sessions for finance officers, closing the loop between what we built and how it got used in the field

Technologies

E-governancePayment GatewaysMobile WalletsTraining

Leadership Experience

Beyond the day job

Leadership

SGA President

San Francisco Bay University

Fremont, CA • January 2024 - December 2024

Represented the student body, managed budgets for events, and partnered with administration to improve campus services and student support programs.

Leadership

Regional Director & Deputy Country Ambassador

Hult Prize and Social Business Programs

Kathmandu, Nepal • March 2019 - February 2022

Led regional-level (multi-country) Hult Prize program, mentored startup teams, and coordinated international social business challenge activities that produced a global finalist team.

7+

Years Experience

$10M+

Daily Transactions

4

Companies

20+

Municipal Deployments