> AIalgo.si: Learn algorithms. Compare AI. Test strategies. > The all-in-one AI lab for builders, students, and analysts. Learn algorithms in an interactive playground, compare AI tools side by side, and backtest trading strategies. ## Pages - [Home](https://aialgo.si/): The all-in-one AI lab — algorithm playground, AI tool comparison, and strategy backtester. - [Demo](https://aialgo.si/demo): Free interactive demo — train a model live, browse the comparison table, run a backtest. No signup. - [Pricing](https://aialgo.si/pricing): Three plans — Explorer (free), Builder ($29/mo), Studio ($99/mo). Yearly billing saves 17%. - [FAQ](https://aialgo.si/faq): 20 answers about the playground, comparison directory, backtester, billing, and privacy. - [Playground](https://aialgo.si/playground): Train k-NN, logistic regression, and decision trees with live decision boundaries (free account required). - [Compare](https://aialgo.si/compare): Directory of 12+ AI tools compared on price, latency, EU hosting, and free tier. - [Strategy](https://aialgo.si/strategy): Educational backtester for MA crossover, RSI, and momentum strategies (free account required). - [Resources](https://aialgo.si/resources): Nine in-depth guides, glossary entries, and benchmarks about ML and AI tools. ## Compare entries - [GPT-4o](https://aialgo.si/compare/openai-gpt-4o): OpenAI's flagship multimodal model with fast, cheap tokens and strong reasoning. Pay per token (from $2.50 / 1M input tokens). EU-hosted: no. - [Claude 3.5 Sonnet](https://aialgo.si/compare/anthropic-claude): Anthropic's balanced workhorse model, strong at coding, analysis, and long documents. Pay per token (from $3 / 1M input tokens). EU-hosted: no. - [Mistral Large](https://aialgo.si/compare/mistral-large): EU-hosted frontier model from Paris — the go-to option for GDPR-sensitive workloads. Pay per token (from $3 / 1M input tokens). EU-hosted: yes. - [Command R+](https://aialgo.si/compare/cohere-command-r-plus): Retrieval-augmented generation specialist with citations built in and EU hosting. Pay per token (from $2.50 / 1M input tokens). EU-hosted: yes. - [Hugging Face](https://aialgo.si/compare/hugging-face): The largest open model hub, with hosted inference and EU-friendly deployment. Free hub + paid inference endpoints. EU-hosted: yes. - [TensorFlow](https://aialgo.si/compare/tensorflow): Google's end-to-end ML framework with mature deployment tooling (TF Serving, TFLite). Free (open source). EU-hosted: no. - [PyTorch](https://aialgo.si/compare/pytorch): The research-first deep learning framework that dominates papers and modern production. Free (open source). EU-hosted: no. - [scikit-learn](https://aialgo.si/compare/scikit-learn): The standard Python library for classical ML: trees, forests, k-NN, regression, clustering. Free (open source). EU-hosted: no. - [XGBoost](https://aialgo.si/compare/xgboost): The gradient-boosting library that wins tabular ML competitions — fast and regularized. Free (open source). EU-hosted: no. - [LangChain](https://aialgo.si/compare/langchain): The most popular framework for wiring LLMs into chains, tools, and retrieval pipelines. Free (open source) + paid LangSmith. EU-hosted: no. - [Pinecone](https://aialgo.si/compare/pinecone): Fully managed vector database for similarity search, with serverless pricing and low latency. Serverless from ~$0.09 / GB-hour. EU-hosted: yes. - [Weaviate](https://aialgo.si/compare/weaviate): Open-source vector database with hybrid search and self-host or EU cloud options. Free (OSS) + managed cloud. EU-hosted: yes. ## Articles - [Decision Trees vs. Random Forests: When to Use Which](https://aialgo.si/resources/decision-trees-vs-random-forests): A single tree is interpretable but fragile; a forest is robust but opaque. Here is a practical decision framework for choosing between them. - [The Complete Guide to AI Model Latency](https://aialgo.si/resources/complete-guide-to-ai-model-latency): Time-to-first-token, p50 vs p95, and why your users perceive your AI product as slow even when the benchmark says otherwise. - [How to Detect Overfitting in Your Backtest](https://aialgo.si/resources/how-to-detect-overfitting-in-your-backtest): A backtest that looks perfect is usually lying. Six diagnostic checks — from out-of-sample splits to parameter sensitivity — that expose curve-fitted strategies. - [EU-Hosted AI APIs: The 2025 Shortlist](https://aialgo.si/resources/eu-hosted-ai-apis-the-2025-shortlist): Which AI providers can actually keep your data inside the EU? A practical shortlist of language models, vector databases, and inference platforms with EU regions. - [k-NN Explained Visually](https://aialgo.si/resources/k-nn-explained-visually): The k-nearest-neighbors algorithm needs no training at all — it just remembers. See why small k means wiggly boundaries and large k means smooth ones. - [Choosing Between Classification and Regression](https://aialgo.si/resources/classification-vs-regression): Will the customer churn, or how much will they spend next month? One is classification, the other regression — and picking the wrong framing dooms the project. - [Sharpe Ratio vs. Sortino Ratio Explained](https://aialgo.si/resources/sharpe-vs-sortino-explained): Both measure return per unit of risk — but Sharpe punishes upside volatility too. Learn when each ratio tells the truth about a strategy. - [What Is AEO and Why It Matters for AI Products](https://aialgo.si/resources/what-is-aeo-and-why-it-matters): Answer Engine Optimization is the practice of making your content quotable by ChatGPT, Perplexity, and Google AI Overviews. Here is how to do it. - [Prompt Engineering Basics for Data Teams](https://aialgo.si/resources/prompt-engineering-basics-for-data-teams): Five prompt patterns that reliably improve LLM outputs on data tasks — extraction, classification, and SQL generation — with before/after examples. ## FAQ - What is AIalgo.si? AIalgo.si is an all-in-one AI lab: an interactive algorithm playground for learning machine learning hands-on, a comparison directory of AI tools and models, and a strategy backtester for testing trading ideas. One account covers all three. - Do I need to know how to code to use it? No. The Playground, Compare directory, and Strategy backtester are all visual and interactive — you pick datasets, algorithms, and parameters with sliders and buttons. The results include plain-language explanations alongside the technical metrics. - Can I try it without an account? Yes. The interactive demo at /demo works with no signup and includes a limited playground, the comparison table, and the backtester. Create a free account to unlock saved experiments, more daily runs, and full datasets. - What can I do with the Playground? You can train classic machine learning algorithms — k-NN, logistic regression, decision trees, k-means, linear regression — on built-in datasets or your own CSV, tune hyperparameters with sliders, and see decision boundaries, loss curves, confusion matrices, and accuracy metrics update live. - Which algorithms are supported? The Playground covers classification (k-NN, logistic regression, decision tree), regression (linear regression, decision tree), and clustering (k-means). The algorithm picker filters by task so you always choose a compatible algorithm. - Can I upload my own data? Yes. Upload a CSV with up to 5,000 rows on paid plans (1,000 rows on the free tier). The first row must contain column names, and you pick which column is the target. Files are processed in your browser and only the resulting metrics are saved. - How many experiments can I run per day? The free Explorer plan includes 5 playground experiments per day. Builder removes the cap entirely, and both plans can save experiments and generate public share links. - What does the Compare directory include? It is a curated directory of AI tools, models, and libraries — currently 12+ entries covering LLM APIs, frameworks, vector databases, and classical ML libraries. Each entry includes pricing, observed latency, EU-hosting status, benchmark numbers, and honest pros and cons. - How is the EU-hosted flag verified? We check for an explicitly documented EU region in the provider's infrastructure docs, review their data processing agreement and subprocessor list, and verify where API requests terminate. A provider must pass all three to be marked EU-hosted. - How current is the data? Entries are reviewed and refreshed at least monthly. Latency numbers are re-measured under identical conditions (same prompt, same region, p50 and p95). Fields we cannot verify are marked "Not disclosed" rather than guessed. - Is the strategy backtester real financial advice? No. The backtester is an educational tool that runs strategies on generated price data to teach how moving-average crossovers, RSI, and momentum strategies behave. It is not financial advice, and backtest results — real or simulated — never guarantee future performance. - Which strategies can I backtest? Version 1 supports moving-average crossover (fast and slow periods), RSI with overbought/oversold thresholds, and time-series momentum with a lookback window. Every run reports total return, Sharpe ratio, Sortino ratio, max drawdown, and win rate. - What data does the backtester use? Version 1 uses deterministic simulated daily price data so results are reproducible and the tool needs no market-data subscription. Real market data integration is on the roadmap for Studio plan users. - Can I cancel anytime? Yes. Cancel from your dashboard in two clicks and you keep access until the end of the current billing period. No phone calls, no retention flows, no dark patterns. - What is the money-back guarantee? Every paid plan carries a 30-day money-back guarantee. If AIalgo.si is not useful to you, email support within 30 days of your first payment and we refund it in full — no questions asked. - Do you offer discounts for students or teams? Students and educators get 50% off the Builder plan with a valid academic email. Teams of 5+ should talk to us about the Studio plan, which includes up to 10 seats and a shared workspace. - Where is my data stored? All user data — accounts, experiments, backtests — is stored in an EU-region Supabase (Postgres) database under GDPR. Uploaded CSV files are processed in your browser and are never stored server-side; only metrics and parameters are saved. - Do you train AI models on my data? No. We never use your experiments, uploaded data, or prompts to train any model. Your saved experiments are private by default and only become public if you explicitly create a share link. - How are the playground algorithms implemented? The demo and playground algorithms (k-NN, logistic regression, decision trees, k-means) are implemented in TypeScript running in your browser, so training is instant and your data never leaves your device. Saved experiments store parameters and metrics, not raw data. - Is there an API? The Studio plan includes API access to the comparison directory (tools, latency, EU-hosting flags, and benchmarks as JSON). Playground and backtesting APIs are on the roadmap. Contact us for early access. ## Legal - [Privacy policy](https://aialgo.si/legal/privacy): What data we collect, where it is stored, and your GDPR rights. - [Terms of service](https://aialgo.si/legal/terms): The rules for using AIalgo.si.