Multi-LLM financial analysis platform
An end-to-end platform that combines market data, automated analysis pipelines, and several language models to produce decision-ready financial reports.
The problem
Financial research often requires analysts to collect fragmented market information, compare signals, and turn it into a consistent report. The objective was to automate the repetitive work without hiding the reasoning behind the result.
Architecture and responsibilities
I designed and built the full platform, from the interactive Next.js interface to Python and Node.js services, asynchronous processing, storage, authentication, and cloud deployment.
- Orchestrated OpenAI, Claude, and Mistral models for sentiment analysis, report generation, and recommendations.
- Built multi-layer ETF and equity filters, automated data pipelines, and real-time scoring.
- Used Celery and Redis to keep long-running analysis outside the request path.
- Secured distributed services with JWT authentication and containerized the system with Docker.
Outcome
The resulting workflow reduced manual research time by more than 70% while preserving a structured, reviewable output. The project demonstrates how I connect AI experimentation to a complete production-oriented product architecture.