Market Knowledge Qualifier

Inspiration

My nightnare: " I need to take an important decision, so I need Information to build my decision on. I am sorrounded by people offering their advises, but most of them dont know what they are talking about. The just give advises to make money no matter what happen to me. I want to create a model that filters noise with a click and create a filtered and certified knowledge for LLMs .

What it does

Market Knowledge Qualifier ingest PDF Papers, Github Repositories, TradingView Strategies, Natural Language Strategies, extracts structured rules, checks whether the strategy is reproducible, verifies that historical data are available, and runs one deterministic out-of-sample test over the rolling one-year window ending today. It then produces a single OOS Reliability Score.

Only records with a score above zero enter the qualified dataset. Incomplete rules, intraday requirements, unavailable data, or subjective parameters receive a score of zero and are archived as rejected. Users can ask questions against qualified records, see citations and limitations, and explicitly opt in before the app uses general AI knowledge.

How I built it

I am not a programmer, I built it by providing a project architecture and then correcting outputs according to my experience in the finance field.

Challenges I ran into

The hardest part was the fornt end: I want to use GPT reasoning capabilities, but at the same time I want for it to work in a restricte environment. Beacuse I trust the reasoning but I dont trust the information available for LLMs to reason about.

Accomplishments I’m proud of

• Finalised the beta version proiect alone with few programming skills

What I learned

There is a new world to be investigated in how address the LLMs in the future

What’s next

To trasform the frontend into a decision support system for traders and investment professionists

Built With

• codex
• gpt-5.6
• next.js
• openai
• python
• react
• typescript
• vitest