HSBC Intelligent Banking ProjectWe built an intelligent banking experience that helps relationship managers and customers make faster decisions using AI-powered insights, automated workflows, and a privacy-first data layer. Our prototype focused on improving onboarding, spending insights, and proactive financial guidance.
Role
Full-Stack Developer
Responsibilities
Responsible for the full-stack development of the web application, covering both client and server-side implementation. Designed the interface, built complete functionalities, and deployed the final product.
Team
- KylieTeam Lead
- SaraUser Research
- HarleenSolution Planning
- Sheng TianSolution Planning
- IsaacMarket Research
Problem Statement
Relationship managers face inefficiencies from fragmented data systems, manual preparation, and disconnected client insights. Valuable time is wasted pulling information from multiple platforms, leading to slower prep and generic client interactions. Without structured reminders or a unified client view, follow-ups are inconsistent and personalized opportunities are often missed. This limits proactive engagement, weakens client trust, and reduces the quality of financial advice. Our challenge was to explore how AI could unify data, streamline workflows, and empower wealth managers to deliver timely, transparent, and personalized financial experiences.
Solution
Our solution reimagined the relationship-manager workflow into an AI-assisted financial intelligence platform — a single, unified dashboard designed to turn fragmented client data into actionable insights. We automated repetitive tasks such as data retrieval and pre-meeting preparation, giving RMs back the time to focus on relationship-building and personalized strategy. Through AI-driven summarization and risk analysis, the system surfaced timely client insights, spending patterns, and proactive recommendations — allowing managers to engage with context, not chaos. We built an end-to-end prototype using Next.js, TypeScript, and Node.js, powered by concept-level vector search and OpenAI-based summarization models to simulate real-time intelligence. The dashboard presented a compliant, privacy-first view of each client's profile, upcoming cash constraints, and suggested actions, blending predictive analytics with explainable recommendations. This created a balance between automation and human expertise — improving efficiency without compromising client trust or data ethics.
Key Flows
- Smart onboarding with guided document analysis and hints
- Unified dashboard with spending insights and anomaly detection
- Goal-based planning with AI-driven nudges and savings suggestions
- Proactive communication assistant for client message drafting
Tech
- Next.js, TypeScript, Tailwind CSS
- Node.js with RESTful APIs
- OpenAI API for summarization and recommendation generation
- Concept-level vector search for insight retrieval
- Privacy-first architecture with data separation and compliance principles
Key Features
Click to see each component
Portfolio Allocation
Visualize asset distribution and investment breakdown
Market Heatmap
Real-time asset class performance monitoring
Product Recommendations
AI-curated suggestions with ESG ratings
Outcome
The prototype demonstrated how AI can elevate the role of relationship managers from reactive advisors to proactive financial partners. Our system helped RMs save up to 45 minutes per client through automated preparation, leading to up to four additional client engagements per day. Smarter follow-ups and reminders ensured that no client was overlooked, while customers benefited from real-time insights and personalized recommendations aligned with their goals and life stages. By merging efficiency with empathy, the solution strengthened client trust and improved engagement quality — directly supporting HSBC's client-centric strategy. The design also outlined a scalable product roadmap, beginning with a pilot rollout in select markets (Singapore, UK) and expanding globally in phased deployment.
Reflection
This hackathon trained me to execute fast while still holding the line on code quality and product clarity. Building the full-stack prototype largely independently meant I had to make high-impact architectural calls early—choosing a clean structure, defining data models quickly, and keeping the implementation maintainable even under pressure. The hardest part was managing scope: there were always more “nice-to-have” features, but I learned to cut aggressively and concentrate on the core user journeys that would best demonstrate value. That MVP discipline helped me ship a working end-to-end experience that communicated a clear story to the judges, rather than a scattered set of half-finished ideas.
Looking back, I’d invest more time in tightening the data visualization layer for readability and insight density, and I’d harden the system with stronger error handling, edge-case coverage, and clearer failure states. The project also reinforced something I think recruiters care about a lot: impact comes from judgment, not just output. While I handled the technical build, my teammates’ research and strategic framing ensured we were solving a real problem with a credible approach, and that alignment made the prototype feel purposeful rather than purely technical. Overall, it strengthened my confidence in shipping under constraints, making trade-offs intentionally, and translating user needs into a product that’s both usable and defensible.