DecideX AI

Inspiration

Today’s world is full of crucial decisions to make, ranging from choosing a career path, building a business, or investing. Unfortunately, most decisions are made based on limited information, individual bias, or excessive information available on the internet.

One thing we realized is that there is a big gap in the market: There is no simple way to make decisions using data, constraints, and intelligent reasoning.

This is what inspired us to develop DecideX AI, which is an AI-based decision copilot.

What it does

DecideX AI assists users in making smart and constraint-based decisions in domains such as: Career decisions Business ideas Investment decisions

Users need to enter: Goal Skills Budget Constraints (e.g., low budget, time constraints)

The system will: Analyze data from the real world (salary data) Understand constraints Use AI agents to assess both positive and negative scenarios Generate a decision with reasoning

It also includes: ROI - Return on Investment Decision Score Confidence Level Risk Assessment

Making decisions not only intelligent but also quantifiable and transparent.

How we built it

We implemented DecideX AI with a 3-layer architecture as follows:

1. Frontend/ UI Layer Built with Streamlit Captures user inputs and displays results with a simple UI
2. Backend/ AI Engine Layer Integrated with Groq API and a powerful LLM model Implemented a multi-agent system with the following roles: Optimist: Identifies factors for success Critic: Identifies risks and challenges Final decision generated by combining: User profile Data insights Constraints Agent reasoning
3. Data Engine/ Logic Layer Manages: Input validation Constraint parsing Salary data analysis ROI and scoring calculation
4. Memory System Stores past decisions in a JSON file Future improvements

Challenges we ran into

1. Combining AI with Real Data LLMs on their own may not be reliable Solution: Using data + rules + AI reasoning
2. Handling Unstructured Constraints Constraints entered by users in the form of NL Solution: Developed a custom constraint parser
3. AI Bias and Hallucination AI may produce answers that are overly optimistic or random Solution: Introduced multi-agent reasoning (optimist and critic)
4. Making AI Explainable Most AI systems operate like a black box Solution: Introduced metrics like ROI, score, confidence, and risk
5. Evaluating Decision Quality Difficult to measure “good decisions” Solution: Developed a custom scoring system

Accomplishments that we're proud of

1. Developed a Multi-Agent AI Decision System
2. Integrated data, AI, and rule-based logic
3. Designed a constraint-aware decision system
4. Provided explainability with measurable metrics
5. Designed a scalable architecture

What we learned

1. How to design real-world AI systems, not just models
2. The importance of the combination of data and AI reasoning
3. How to make AI systems transparent and explainable
4. How to design a modular architecture (UI, Backend, Logic)
5. How to effectively handle edge cases and user constraints

What's next for DecideX AI

1. Technical Enhancements Real-time data sets (Job Market, Industry Trends) Advanced NLP for better understanding of constraints Personalized recommendations through memory
2. Product Expansion Career counseling platform for students AI assistant for startup founders Smart investment advisor
3. Future Vision To evolve DecideX AI into a universal decision intelligence platform that enables people to make better life decisions with confidence and clarity

Built With

• csv
• github
• groq
• json
• numpy
• pandas
• python
• regrex
• streamlit