Market Research Assistant

Problem statement

Private investment research is often scattered across spreadsheets, news tabs and ad hoc notes without a consistent way to compare companies, document thesis changes or review weighted scoring.

Engineering objective

1. Mapped watchlist, company dashboard, news and research journal workflows.
2. Defined transparent weighted scoring rather than opaque model output.
3. Scoped server-side API routes so secrets remain off the client.
4. Planned PostgreSQL schema with local JSON fallback for development.

Solution

A Next.js research workspace for watchlist management, company dashboards, financial metrics, news aggregation, research notes and DeepSeek-assisted analysis with deterministic mock behaviour when credentials are unavailable.

• Watchlist CRUD and per-company overview pages.
• News aggregation and AI classification via server routes.
• Weighted score breakdown with chart visualisation.
• Trend snapshots recorded on each analysis run.

How it works

Next.js App Router, TypeScript, Tailwind CSS v4.

Recharts, Drizzle PostgreSQL schema, local JSON store fallback.

DeepSeek API via OpenAI-compatible server routes.

Optional Yahoo Finance, Alpha Vantage, Finnhub and SEC EDGAR integrations.

Controls and constraints

• API keys handled only in server-side routes.
• Explicit disclaimer that outputs are not financial advice.
• Mock deterministic analysis when AI credentials are unavailable.

Key capabilities

• Dashboard, watchlist and company detail pages.
• Research journal and trend snapshot storage.
• Server API routes for analysis, news, financials and settings.
• PostgreSQL schema and scoring calculator.

Product and system views

A structured research workspace combining market data, news signals and explainable AI-assisted analysis.

What the system enables

• Demonstrates explainable AI-assisted research rather than black-box recommendations.
• Shows how to structure a private research tool with honest limitations.

Validation and testing

• Weighted scoring system documented with explicit factor weights.
• Mock analysis path for UI exploration without live credentials.

Limitations

• Not financial advice or a trading system.
• Default store uses local JSON unless PostgreSQL is configured.
• Free market data integrations are delayed and rate-limited.