CoachInvest

Making professional‑grade investing knowledge accessible to everyone: CoachInvest is a platform designed to transform active portfolio management by providing actionable, data-driven guidance.

Comment

Project Snapshot: Democratizing Active Portfolio Management

Vision – Make sophisticated portfolio‑building accessible to every Hong Kong investor, while fostering a culture of learning and self‑reflection on behavioral biases.

1. Problem Statement

The "What": Hong Kong has an incredibly active retail investing population

Context Detail
Low entry barrier Anyone can buy a single fractional share for HK$100.
High participation rate & account penetration Over 2.7 million individual investor accounts
Ease of execution Over 90% of retail trades are executed online or via mobile apps

The "So What": Evidence of the knowledge gap

Issues Detail
Performance chasing & high volatility Retail investors accounted for a significant portion of highly speculative momentum trading e.g. "meme stock" episodes
Data on Financial Literacy Local Survey Data found that only ~40% of respondents could pass a basic quiz on financial concepts like inflation, compound interest, and risk diversification.
Derivatives dominance for retail Hong Kong is the world's largest market for retail-driven derivatives, which are complex, leveraged, and high-risk products designed for short-term speculation, not long-term wealth building.

The problem is not a lack of access, but a deficit in the strategic literacy required to use that access effectively for long-term financial health. The ecosystem provides the "how" to trade (apps, platforms) but fails to provide the trusted "why" and "what for" that defines professional investment practice.

Democratizing investing knowledge is not a zero-sum game; it creates a larger, more resilient, and innovative financial ecosystem.

We decided to solve the knowledge gap directly: instead of a generic course, provide real‑time, active portfolio‑management guidance that can be everyone's trusted coach.

2. Killing Features

  • The “Context Provider” – Portfolio Insights with a “So What?” filter. Transforms complex, disjointed portfolio data into a coherent, executive‑level summary.
  • The “Tactical Assistant” – LLM‑powered Buy/Sell/Sizing Signals. Contextual, data‑driven signals integrated with the user's portfolio and historical investment behavior.
  • The “Devil’s Advocate” – Risk mitigator engine to test out ideas. An interactive module that a user triggers before a buy/sell decision. The AI coaches and challenges the user's investment thesis with insights.
  • The “Report Card” – Performance Analytics & Benchmarking. A personalized dashboard that compares the user's simulated or linked portfolio performance against benchmarks.

3. Development Sprint (Week 1–2)

Day Activity Tools
Day 1‑3 Prototype core logic on FastAPI + GitHub Copilot Python, FastAPI (we dropped), Copilot
Day 4‑5 Integrate market‑data pipelines (Yahoo Finance) pandas, yfinance
Day 6‑7 Build a lightweight web UI prototype Kiro, React
Day 8‑9 Migrate API layer to AWS Lambda for serverless scaling AWS Lambda, API Gateway
Day 10‑11 Hook Amazon Bedrock (Nova Premium) as LLM engine & perform prompt tuning for portfolio explanations Bedrock Nova Premium, custom prompts
Day 12‑14 Deploy static front‑end on S3 + CloudFront; set up monitoring S3, CloudFront, CloudWatch

Why Bedrock Nova Premium?
1️⃣ It offers the best cost‑efficiency for real‑time language models.
2️⃣ It can process niche domain knowledge like portfolio management. We tried other models, the performance was not as great as Nova Premium.

4. Architecture Overview

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   React Frontend│    │   API Gateway   │    │ Lambda Functions│
│   (TypeScript)  │◄──►│   (REST API)    │◄──►│   (Python)      │
└─────────────────┘    └─────────────────┘    └─────────────────┘
         │                                               │
         ▼                                               ▼
┌─────────────────┐                            ┌─────────────────┐
│   S3 Hosting    │                            │   DynamoDB      │
│  (Static Site)  │                            │  (NoSQL Data)   │
└─────────────────┘                            └─────────────────┘

Components

Layer Responsibility AWS Service
Frontend UI/UX, user interactions S3 (static hosting) + CloudFront (CDN)
API Gateway Expose REST endpoints, throttling API Gateway
Lambda Execute portfolio logic, call Bedrock Lambda
Bedrock Nova Premium Generate natural‑language trade insights and bias explanations Bedrock
CloudWatch Log all API calls, monitor latency & errors CloudWatch

5. Why This Stack?

  • Serverless (Lambda) → Zero‑ops scaling; pay only for execution time.
  • Bedrock Nova Premium → Low‑latency LLM inference with the best performance.
  • S3 + CloudFront → Fast, globally available front‑end delivery.
  • DynamoDB → Scalable, low‑latency storage for portfolio and market data —ensuring fast reads/writes without provisioning capacity.
  • CloudWatch → Full observability of every request and LLM inference, essential for auditing.

Bottom line: This stack lets us ship an MVP in weeks, not months, while keeping the codebase clean, maintainable, and ready for future expansion.

6. Next Steps

  1. Feature Gap – Current release covers < 30 % of our vision.
  2. Expand Active Portfolio Management – Add multi‑asset classes, dynamic rebalancing rules, and risk‑adjusted performance metrics.
  3. Build More Agents – Deploy specialized LLM agents for goal setting, tax planning, and ESG screening to deliver accurate, customized coaching.
  4. Real‑time Market Data – Integrate streaming feeds (HKEX WebSocket API) so the system reacts instantly to price moves.
  5. Multi‑User & Multi‑Portfolio Support – Allow users to create several portfolios under one account, with isolated data and dashboards.
  6. Custom Benchmark Creation – Let the LLM guide users in building personalized benchmarks (e.g., “A portfolio of HK 10‑day momentum plus 30% bonds”) that align with their risk appetite.

If we win the hackathon we’ll have a fully fledged, data‑driven active‑management platform that not only teaches but does the heavy lifting for retail investors—turning intuition into informed action.

Share this project:

Updates