Projects

InvestIQ

Full-stack AI investment research platform aggregating 9 data sources with Gemini-powered analysis

Production
90%
Time Reduction
9
Data Sources
$200K+
Annual Savings
Finance & AI
React, Python, FastAPI, Gemini
Individual Project
CHALLENGE

Research Inefficiency

Investment research is time-consuming and fragmented across multiple data sources. Analysts spend hours manually aggregating information, extracting financial data from unstructured sources, and performing technical analysis.

15+ hours per comprehensive research report
Data scattered across 9+ different sources
Manual extraction of private company financials from news
SOLUTION

AI-Powered Research Platform

A comprehensive platform that automatically aggregates data from 9 sources, uses Gemini AI to extract structured financials from news articles, performs technical analysis, and generates investment-grade reports in minutes.

9-source data aggregation with intelligent caching
Gemini-powered financial extraction from unstructured text
Automated technical analysis and sentiment scoring

Business Impact

0%
Time Reduction
From 15 hours to 10 minutes per report
$0K+
Annual Savings
Cost savings with zero operating costs
0
Data Sources
Unified access to financial, news, and market data

Technical Architecture

Frontend
React
TypeScript
API Layer
FastAPI
Python
AI & ML
Gemini AI
FinBERT
Infrastructure
Docker
HuggingFace

Framework & Approach

Multi-Agent AI System with 4 specialized agents: Data Gathering → Analysis → Synthesis → Report Generation with graceful degradation so system works even if AI fails

1

Phase 1: Foundation - API integration and data pipeline (9 sources including yfinance, NewsAPI, GitHub)

2

Phase 2: AI Agents - Gemini + FinBERT integration for financial extraction and sentiment analysis

3

Phase 3: Advanced Capabilities - Private company financial extraction from news, Hidden Signals analysis

4

Phase 4: Deployment - Docker optimization, HuggingFace deployment, multi-stage builds

5

Phase 5: Production - Testing with real companies, validation, cost optimization to $0/month

What This Project Demonstrates

Transferable skills and capabilities beyond the technical implementation

Problem Decomposition

Identified that investment research bottleneck was data aggregation across 9 fragmented sources taking 15 hours. Designed multi-agent architecture where each agent handles specific data source efficiently.

Problem AnalysisSystem ArchitectureAgent Design

AI Reliability Engineering

Built JSON extraction with regex fallback, retry logic with exponential backoff, graceful degradation so reports work without AI. Production systems need resilience patterns.

Error HandlingFault ToleranceProduction Readiness

Deployment Constraints

Hit Vercel 50MB serverless limit, pivoted to unified Docker deployment on HuggingFace. Understanding platform trade-offs crucial for production decisions.

Platform SelectionCloud ArchitectureAdaptability

Data Quality Transparency

Implemented explicit confidence levels (HIGH/MEDIUM/LOW) for different data sources. Free APIs have real constraints - designed around them with honesty rather than hiding limitations.

Data IntegrityUser TrustHonest Communication