Stocki.ai

AI-powered Investment Assistant: backtesting, insights, and community in one place.

Responsibility

Functionalities

AI-native platform. We are committed to deeply integrating the investment research process with AI services to provide a convenient, reliable and accurate investment research experience.

• Stock Agents: AI Agents dedicated to individual stocks, based on the financial database of private equity funds, responds with the latest and most accurate data, and supports backtesting, portfolio construction, etc
• Community: A platform for stock and investment exchanges, where you can get the latest cutting-edge information, communicate with senior investment bankers, and quickly share your insights with the help of AI.
• Earnings & Watchlist: Check which company has released the latest financial report, quickly build your investment portfolio through Stocki AI, verify your investment portfolio through backtesting, and continuously track the latest situation of your investment portfolio.
• Library: Timely add the investment research data and the analysis results of the agent to the knowledge base for easy review and better analysis.

My Contribution

• Led frontend development for the financial AI research product Stocki, driving core module design and delivery from 0–1 MVP to v1.8. Managed a team of three engineers and introduced multiple engineering and UX optimizations.
• Delivered key product modules through cross-functional collaboration (UI, engineering, QA), including AI agent chat, general backtesting, knowledge base, user profile, and stock search.
• Built proto4all, a middleware that generates APIs, implementations, mocks, and OpenAPI 3.1 specs directly from .proto, enabling API-First and SSOT practices. Significantly decoupled frontend/backend/test workflows and improved overall development efficiency by ~30%.
• Designed a natural language-driven general backtesting system, integrating financial research workflows (screening, factor investing, strategy building) into conversational AI through intent recognition and guided interaction.
• Architected and optimized a frontend–backend–LLM streaming system based on HTTP streams, introducing event-driven processing, structured data rendering, chain-of-thought handling, and heartbeat mechanisms. Improved AI interaction development efficiency by ~80%.
• Proposed and implemented a frontend–backend–blockchain consistency model, enabling stateless frontend design, supporting decentralized architecture, and achieving both system-level consistency and eventual consistency.