Work

Realtime learning pipeline

Lead the core learning pipeline, the engine where each per-user agent makes its personalization decisions. Moved it from a batch setup to realtime so an agent can act on what a user just did, not on what they did hours ago.

Applied AI for the org

Head Applied AI across the company: raising AI literacy, designing agentic workflows, and building internal and customer-facing tools. Ran POCs on how agentic workflows speed up product onboarding and usage, GTM, and content creation. As a MarTech company our customers live on content, so agent-assisted content generation is a direct lever on their success.

We run always-on agent harnesses that act on internal-tool feedback autonomously, plus a zero-bug policy enforced by an agent loop running around the clock with browser and vision capabilities.

Core transport booking and order systems

Owned the engineering surface for how customers booked cab rides, cancelled, rebooked, changed ride types, and moved through the transport funnel. Re-architected the core booking engine to be far more expandable, broadened payment options including split and pay, and added multi-select so riders could book more than one ride type at once. Built haggle, letting riders negotiate the fare directly with drivers. Designed the booking transaction and order-management systems that handled very different order types across regions: scheduled rides, prepaid four-wheeler trips, and multi-modal journeys.

We shipped the first-ever multi-modal flow, letting a single booking combine a two-wheeler ride, public transport, and a four-wheeler ride end to end.

Personalization as a growth lever

Turned personalization into a real growth lever for the mobility business. Built a personalized homepage that genuinely targeted each user, speaking to their needs, value systems, and preferences rather than running as a blunt experiment. Reordered the services each rider saw through intelligent personalization and recommendation systems, so the most relevant way to move came first.

Real personalization is not an A/B test you leave running. It is a model of what each rider actually values and how they like to move.

Leading mobility engineering

Led engineering for the core mobility business, where the responsibility was not a single feature lane but the transport offering itself: millions of bookings daily, hourly spikes, order state, payments, booking reliability, rider experience, and the roadmap behind how the business moved. Owned product-engineering direction with business leaders, managed engineers with diverse functional focus, and coached people across the organization on execution and career development.

Platform engineering and team leadership

Managed the team responsible for creating and enhancing the live education platform to cover all aspects of education during the COVID lockdown period. Day-to-day included alignment with business and engineering teams, mentoring team members on career development, and shipping committed deliverables on time.

Communications systems for 15+ products

Built the communications platform responsible for reaching millions of users across more than 15 Gojek products. Each product had its own transactions, lifecycle journeys, operational alerts, and marketing campaigns, so the real scale was millions of users multiplied across many independent product lines. The platform handled traditional channels like push notifications, SMS, and email, alongside owned in-app channels: inbox, B2C and C2C chat, and a stories-style surface that let marketing teams ship richer product updates inside the app.

We built this because CleverTap could not carry the volume or product-specific complexity. That platform later became the foundation over which Gojek’s in-house CleverTap replacement was built.

Guardrails, masked calls, and core product work

Built notification guardrails so product teams could avoid bombarding users, using a rules system that was the most practical governance model we could get approved at the time. Built number masking to protect driver and customer identities when they needed to talk. That path moved me from platform systems into core product work and into the quiet effort of bringing Gojek to Singapore.

I later got to build a much smarter version of the notification-overload problem at Aampe: less rules theater, more learning system.

Ledger and reporting engine

Developed a ledger and report generation engine that consumed, consolidated, and calculated investments and expenses, automatically generating reports for investors in multi-billion dollar hedge funds. Designed a self-service screen for the reconciliation framework, reducing manual dev intervention by 30%.

Generic Reconciliation Framework

Built the reconciliation framework on a cloud-based post-trading platform serving alternative asset managers like Blackstone, JPM, and D.E. Shaw. The engine reconciled at configurable granularity with pluggable data sources, delta processing, and application metrics.

A single mismatched penny in a reconciliation run can signal a systemic issue worth investigating. You learn to take small discrepancies seriously.

Workshop curriculum and mentorship

Mentored students in enterprise tools and technologies like Git, React, React Native, Django, and Spring MVC through intensive summer and winter camps. Kept coming back to contribute across multiple seasons while working full-time elsewhere.

Explaining recursion to a 14-year-old is a different challenge than implementing it, and both make you better.

Fitness-tech platform

One of the early attempts at building a product from scratch. The platform brought together different entities in the fitness industry to interact, collaborate, and grow. Learned to ship fast and iterate on real user feedback.

ssh_scan REST API and architecture

Built the initial REST API structure, added SSL support, and planned the job queue and scaling architecture using thread pools. Also took on a mentor role, introducing juniors from college to open source development.