Round 1 Winners!
Proof of Usefulness Report

Tinybrain++

Analysis completed on 4/21/2026

+43
Proof of Usefulness Score
You're In Business

Tinybrain++ offers a compelling theoretical solution for fast, explainable, and lightweight machine learning at the edge. However, as a pre-commercial, pre-launch project validated primarily on synthetic data with no active user base or revenue, it falls into the minimal traction tier. Its strong market timing and technical relevance provide a solid foundation, but significant real-world validation is required to achieve a higher PoU score.

View All Reports

Score Breakdown

Real World Utility+15.0
Audience Reach Impact+1.0
Technical Innovation+7.5
Evidence Of Traction+2.5
Market Timing Relevance+7.0
Functional Completeness+4.0
Subtotal+37
Usefulness Multiplierx1.15
Final Score+43

Project Details

Description
TinyBrain++ is a ~5 MB white-box ML engine for fraud, churn, anomaly, stock, crypto & medical devices. Tested on fraud (30K txns, 0.45% fraud rate): 97.78% recall, 99.25% precision, 0.0071ms inference on CPU. Comparatively smaller than XGBoost, fully explainable (no black box). Alternative to neural networks.
Audience Reach
Pre-launch. Projected reach: 140,391 transactions/second processing capacity. Target audience includes banks (fraud detection), hospitals (medical wearables), telecom (customer churn), and manufacturing (predictive maintenance). Currently in development stage with validation on synthetic fraud data (30K test samples, 97.78% recall).
Target Users
Banks and fintechs (fraud detection), medical device companies (wearable monitors), telecom operators (churn prediction), manufacturers (anomaly detection), crypto exchanges (transaction monitoring), and regulators (explainable AI compliance). No notable customers yet – pre-commercial stage.
Technologies
Other, Python, NumPy, LightGBM, Scikit-learn, Tensorly, XGBoost, imbalanced-learn (SMOTE), FastAPI (for API deployment), Joblib (model serialization).
Traction Evidence
1. HackerNoon article: 107 views, 9 reactions in first 2 days https://hackernoon.com/tinybrain-a-compact-interpretable-alternative-to-black-box-ai 2. Contacted SEON (fraud detection company) – awaiting response 3. Contacted H2O.ai – awaiting response 4. Accepted into The Hong Kong Polytechnic University (Computing & AI, Sept 2026) 5. Preparing HKSTP incubation application for startup funding 6.Polyu KTEO

Algorithm Insights

Market Position
Growing utility with room for optimization
User Engagement
Documented reach suggests active user community
Technical Stack
Modern tech stack aligned with sponsor technologies

Recommendations to Increase Usefulness Score

Document User Growth

Provide specific metrics on user acquisition and retention rates

Showcase Revenue Model

Detail sustainable monetization strategy and current revenue streams

Expand Evidence Base

Include testimonials, case studies, and third-party validation

Technical Roadmap

Share development milestones and feature completion timeline