CapitalAI
Our target users are beginner investors, university students, and young professionals who want to invest but lack the financial knowledge and analytical tools to make informed decisions. Today, they struggle to interpret financial reports, understand market indicators, compare investment opportunities, and assess potential risks. Existing financial platforms often present complex data that is difficult for non-experts to understand. As a result, users may make decisions based on incomplete information, emotional reactions, or unreliable online advice. Investor AI solves this problem by transforming complex financial data into clear, accessible, and personalized investment insights.
Preliminary score
- Value for the user15.0 / 25
A specific problem and a clear outcome from solving it.
- Prototype and use of AI16.0 / 30
A working core scenario, and what AI actually contributes to the result.
- Quality testing11.0 / 20
Tests, examples of failures, and comparison with the current approach.
- Feasibility9.0 / 15
Data requirements, running costs and a clear next step.
- Originality5.5 / 10
A meaningful difference from common solutions.
What the judges wrote
- GPT-6 AstraPreliminary58/100
- Strengths
- Your separation of financial calculations from AI explanations makes the numerical reasoning inspectable. Your README documents concrete fixes, including incorrectly matched spreadsheet rows, misleading gross-return optimization, and invalid EBITDA handling. Comparing investment amounts through IRR, risk constraints, and opportunity-cost NPV gives your project a clearer decision-support purpose than a generic financial chatbot.
- To improve
- Show a reproducible upload-to-result run with an actual AI explanation, demonstrating its benefit beyond your rule-based fallback. Provide expected-versus-actual results for the documented tests and compare recommendations with an analyst's spreadsheet, including sensitivity to assumptions. Reconcile the deck's stated 12% EBITDA payout with the README's investment-scaled payout formula, and validate the model on expert-reviewed company cases rather than synthetic data alone. Choose one initial audience—beginner investors or professional analysts—and measure task time, user comprehension, and per-analysis cost in your pilot.
- Claude Opus 5.5Preliminary55/100
- Strengths
- Your engine is transparent. Altman Z, CAGR-based EBITDA growth, leverage against Damodaran sector limits, a DSCR-style default check, and 4,000 Monte Carlo scenarios per amount are all spelled out, mapped to functions in engine.py, and their assumptions (12% payout, 40% recovery) are clearly labelled. Choosing the amount by NPV against an alternative return, rather than maximum total payout, is a thoughtful design. Your README shows you caught and fixed that exact flaw, along with others such as using common random numbers across amounts. The 400K vs 430K example on the demo slide makes the trade-off concrete for an investor.
- To improve
- Your target user is inconsistent: the submission says beginner investors and students, while the deck says investors and financial analysts. The tool also needs company financials, so pick one user and quantify their current pain (time or cost of doing this by hand). The test slide lists six cases that all 'passed' with no concrete failure examples. The written 'what broke' section is generic, and you never compare the tool's recommendation with an expert's or a spreadsheet's answer. The LLM is optional and only explains results, so show what GPT-4o mini adds over the rule-based explanations (e.g. accuracy checks of its explanations) and estimate a per-analysis running cost.
The team
Nijat Babayev
business
Arzu Ismayilzade
data
Read SamadovLead
business