cool things i’ve worked on
Over the course of my professional life, i’ve had the pleasure of working on a variety of ‘cool things’ if you will. these range from marketing projects for a startup swimwear company to analysing credit card fraud with hsbc. you can take a look at a few of the projects here:
corporate projects (HSBC)
digital verification
Opening a bank account can be quite the process, and if you had to go through this the traditional way, I’m sure you can attest.
There were many factors involved in launching the new digital identity verification solution for the HSBC HK app, including vendor selection, testing false positive and negative rates of facial scans and id scans, creating educational materials, and conducting user acceptance testing during app off-peak hours. We tested under a variety of conditions to validate the vendor quality and in a way try to trip the system out to see how easy it would be to bypass it for fraud.
For a first rotation, this was an incredibly eye-opening experience into how thoroughly all aspects of a new rollout are vetted and ensure they pass through rigorous risk and legal requirements. We spoke with project leads in Poland and the UK, backend developers in China, and business heads in Hong Kong. The scale at which these projects are delivered is something I’m eager to continue contributing to and advancing.
Alongside this project, I also conducted ad-hoc analysis on user drop-on and drop-off rates to ensure app stability and monitor the effects of system outages on key metrics. A particularly interesting analysis involved converting hourly-level data to minute-level data, which required experimentation with Google Looker Studio. This analysis revealed when most users accessed the app throughout the day, which directly informed our advertising campaigns, in-app messaging, and broader engagement strategies.
CREDIT CARD FRAUD ANALYSIS
coming soon…
GEN AI COMPLAINT AUTOMATION
My second rotation focused on implementing generative AI in the customer experience and relations department. I analyzed HSBC’s complaint management process, where handlers investigate complaints and draft formal responses, and identified a dual-input system (complaint category + summary) that could feed an AI model to generate standardized response drafts. By studying 20–30 historical complaint responses, I mapped which human inputs would most effectively trigger response generation for different complaint types.
we collaborated with the Microsoft Copilot team to build an AI-integrated complaint management solution using a Human-in-the-Loop validation framework, allowing handlers to input complaint details and receive AI-generated responses which were personalized and checked before sending. This approach augmented handler capabilities rather than replacing them, maintaining compliance with HKMA standards while eliminating manual drafting time.
To assess impact, I led an 3-week validation study with 30+ handlers, iterating on AI prompts and establishing accuracy benchmarks.i also created the training materials, hosting sessions for complaint handlers to understand the systems capabilities and usage constraints.
Til date, this system is being used by frontline staff at hsbc, making what used to be an extremely manual and time-consuming task, into a far more streamlined one. after the first and second round of testing, handlers reported a near 35% reduction in time spent per complaint case from 1 hr 15 mins + to under 45 minutes.
transaction banking T&C
coming soon…
marketing internship with rii swim
coming soon…
Esha Mahbubani 