Back to all cases
CASE STUDY / AI & Data Science
Stock Multi-Agent Financial Analyzer
FinTech, Investment Research & Decision SupportFinancial analysis platform combining an LLM Gateway with Playwright browser automation. Multi-agent collaboration (news / finance / trader / data) with Safety Guard and full artifact traceability.
Delivery Details
VICI is a production-grade LLM Gateway + multi-agent financial analysis system. The Gateway layer uses a Provider abstraction (strategy pattern) supporting one-click switching between OpenAI / Claude / Mock, with a built-in 5-layer Safety Guard (input validation → prompt scan → provider allowlist → output redaction → audit log). Four specialized agents (News / Finance / Trader / YFinance) collaborate via Playwright headless browser to scrape stock data and news, then LLM generates investment reports with sentiment analysis. Each run produces full artifacts (report.md / slides.pdf / run.json / trace.zip / SHA256 checksums), with deterministic dry-run for offline testing.
Industry Context
Investment research combines market data, fundamentals, news, and technical indicators with different sources and update cycles, while every conclusion must retain its evidence and reasoning trail. Multi-agent systems can separate specialist tasks, but financial use also requires safety controls, replayable workflows, and complete auditability.
Business Problem
Investment research workflows were fragmented, with weak traceability across sources, reasoning, and reports.
What I Built
Connected news, finance, trader, and data agents through an LLM Gateway with safety controls.
Business Impact
Packaged analysis, audit logs, and replayable artifacts into a verifiable research workflow.
For career conversations, advisory work, or a concrete Data & AI project, feel free to get in touch.