SYLA Analytics

OVERVIEW

Overview

SYLA Analytics is an AI-powered data platform that automates chart generation, provides natural-language forecasting and offers file conversion tools. It was built with React, FastAPI and OpenAI models to remove the manual work of turning a spreadsheet into charts and a report.

The first version was built and demonstrated end to end. SYLA is now being rebuilt as GOLIDE's first SaaS product, and this page documents the original build.

The problem

• Popular analysis tools require manual chart creation for every dataset.
• Reports cannot be generated automatically from raw data.
• Forecasting requires specialised statistical knowledge.
• Enterprise analytics tools are expensive and slow to set up.
• File-format conversion is a repetitive, time-consuming chore.

Methods and tools

Frontend

React 18, Chart.js, TailwindCSS and Vite.

Backend

Python 3.11+, FastAPI, Pandas and Prophet.

AI models

OpenAI GPT-4o-mini for natural-language scenarios, combined with Prophet time-series models.

Database

PostgreSQL with SQLAlchemy ORM.

Authentication

JWT tokens and Google OAuth 2.0.

Deployment

AWS: EC2, IAM, VPC and Aurora/RDS.

How it works

• Upload: drag and drop a CSV or Excel file of up to 10 MB.
• Auto-process: SYLA cleans the data, detects column types and generates summary statistics.
• Visualise: bar, line, pie, scatter and area charts are generated automatically, with customisable options.
• Forecast: describe a scenario in plain English, such as '15% marketing increase next quarter', and get a forecast with 95% confidence intervals from a hybrid GPT + Prophet model.
• Convert: move between PDF, Excel, CSV and Word, merge up to 15 PDFs, and compress files at three levels.
• Export: download charts as PNG, SVG or PDF, or integrate through a REST API with OpenAPI documentation.

Applications

• Business analytics: sales forecasting, revenue trends and customer analysis.
• Healthcare: patient data visualisation, epidemic modelling and resource planning.
• Finance: portfolio analysis and risk assessment.
• E-commerce: demand forecasting, inventory planning and conversion tracking.
• Education: student performance tracking and enrolment predictions.

Results and rebuild

• Reported for the original build: chart creation time cut from about 5 minutes to about 15 seconds.
• Reported for the original build: 85-95% forecast accuracy for short-term predictions (1-12 periods ahead).
• Files of up to 10 MB with 100,000+ rows handled efficiently.
• The rebuild focuses the product on visualisation, forecasting and file intelligence in one workspace.

Explore the original work

• Backend source code
• Original project page
• Demo on YouTube