Inspiration
Crypto trading decisions are often made under extreme uncertainty, emotional pressure, and incomplete information. Most tools focus on automation or raw signals, which can amplify mistakes instead of reducing them.
This project was inspired by a core question: How can AI help humans make better decisions without replacing human judgment?
Instead of building another trading bot, I wanted to create an AI system that supports thinking, explains its reasoning, and encourages safer decision-making in high-risk environments.
What it does
Crypto Decision Copilot is a human-in-the-loop AI system designed to assist traders in volatile crypto markets.
It analyzes:
📊 Market data and technical indicators across multiple timeframes
📰 Crypto news and sentiment as external context
⚠️ Risk constraints and uncertainty factors
It produces:
A trade recommendation (BUY / SELL / HOLD)
A confidence score reflecting uncertainty
A clear explanation of the reasoning process
A risk-aware trade plan
The AI does not execute trades. Final decisions always remain with the user.
How we built it
The system is implemented as a modular decision architecture, inspired by real-world AI systems used in high-stakes domains.
Core components include:
Market Analysis Module Processes OHLCV data and computes technical indicators across timeframes to capture both short-term momentum and broader market structure.
News Analysis Module Incorporates recent crypto news and sentiment to detect external risk factors that price data alone may miss.
Decision Agent Synthesizes market signals, news context, and predefined risk policies into a structured, explainable decision output.
Human-in-the-Loop Layer Ensures the AI provides guidance and reasoning while preserving human oversight and control.
The design prioritizes interpretability, safety, and extensibility over opaque optimization.
Challenges we ran into
Resolving conflicting signals between technical indicators and news sentiment
Designing risk logic that is conservative without becoming unusable
Preventing overconfident AI recommendations in uncertain market conditions
Keeping the system modular while increasing decision complexity
These challenges directly shaped the system’s emphasis on transparency and risk awareness.
Accomplishments that we're proud of
Built a reasoning-based AI decision system, not just a signal generator
Designed a clear human-in-the-loop architecture for high-risk decisions
Integrated multiple data sources into a coherent, explainable output
Created a foundation that can be extended to other high-stakes domains
Most importantly, the system consistently explains why a decision is suggested.
What we learned
This project reinforced several key lessons:
AI systems are most valuable when they augment human judgment, not replace it
Explainability and confidence calibration are critical in high-risk domains
Decision quality matters more than raw prediction accuracy
Responsible AI design requires intentional constraints and transparency
Building this project deepened my understanding of human-centered AI.
What's next for Crypto Decision Copilot
Future improvements include:
Deeper integration with Gemini models for reasoning and contextual analysis
More advanced confidence calibration and uncertainty estimation
Support for additional assets and market regimes
A richer UI for interactive decision exploration
The long-term vision is to evolve Crypto Decision Copilot into a trustworthy AI reasoning partner for complex, high-stakes decisions.
Log in or sign up for Devpost to join the conversation.