Golden Bear: Wealth Management Agents

Inspiration

Managing personal finances today feels like juggling a thousand disconnected pieces–investments, taxes, risk, goals, market news, and life changes. Traditional tools (Mint, YNAB, EveryDollar) treat everyone the same, while real financial advisors are expensive and operate without real-time context. We wanted to build something that feels like having a full wealth-management team dedicated to you, but powered by autonomous agents and cutting-edge memory systems. Our inspiration was simple: What if expert-level financial planning was accessible to everyone, 24/7, no matter their income or background?

What it does

Introducing GoldenBear: Wealth Management Agents, a revolutionary agent-orchestrated, memory-augmented financial advisory platform. GoldenBear does three big things:

Understands the user deeply using Fastino’s profile builder (income, age, risk tolerance, goals, tax info, portfolio breakdown)
Analyzes and plans using a team of specialized Google ADK agents coordinated by the custom implemented Financial Advisory Orchestrator (agents: portfolio management, risk assessment, tax optimization, planning, compliance, and live market research).
Intelligent memory: features episodic, semantic, and procedural memory–creating continuity like a real expert financial advisor who learns over time. Powered by GeminiAIClient for high-level reasoning and Linkup for real-time market intelligence, it generates dynamic, personalized, goal-aligned financial guidance that evolves in real-time with each new financial event.
Welcome to the future of personalized agentic finance advising.

How we built it

We built GoldenBear by using Fastino to generate a structured user profile and portfolio, then connected it to a Gemini-powered multi-agent system that includes portfolio, tax, risk, research, planning, and compliance agents. The Financial Advisory Orchestrator coordinates these agents while Linkup supplies live market data. Finally, we layered in episodic, semantic, and procedural memory so the system can track recent events, retain financial knowledge, and repeat workflows intelligently.

Challenges we ran into

Agentic systems require orchestration, not just intelligence—the meta-controller matters as much as the agents.
Financial reasoning benefits massively from memory structures, especially episodic memory for “recent market events” and procedural memory for repeatable workflows like rebalancing.
Good user modeling is foundational—risk tolerance, goals, and tax info deeply transform the advice an AI can give.
Real-time external research (via Linkup) is crucial for relevance; static models alone can’t keep up with markets.
AI financial planning is less about answers and more about alignment—matching suggestions to user goals, constraints, and compliance rules.

Accomplishments that we're proud of

A fully functioning multi-agent financial advisory system orchestrated autonomously through Gemini.
A three-layer memory architecture that enables context retention, historical reasoning, and reusable procedures—rare in hackathon projects.
Dynamic investment planning that adapts to the user’s goals, risk tolerance, and timeline in real time.

What we learned

Agentic systems require orchestration, not just intelligence--the meta-controller matters as much as the agents.
Financial reasoning benefits massively from memory structures, especially episodic memory for “recent market events” and procedural memory for repeatable workflows like rebalancing.
Good user modeling is foundational—risk tolerance, goals, and tax info deeply transform the advice an AI can give.
Real-time external research (via Linkup) is crucial for relevance; static models alone can’t keep up with markets.
AI financial planning is less about answers and more about alignment--matching suggestions to user goals, constraints, and compliance rules.

What's next for Golden Bear: Wealth Management Agents

Full automation of portfolio updates with user-approved rebalancing and tax-loss harvesting.
Natural-language financial coaching, turning the system into a conversational planner who teaches as it advises.
Expansion of memory, enabling multi-month trends and personalized financial storytelling.
Mobile deployment, making GoldenBear a pocket-sized wealth advisor available anywhere.