Bet365 Sports Automation Platform
Bet365 is a real-time sports operations platform that ingests Telegram publisher feeds, parses structured event signals, and executes actions on the Bet365 platform with full auditability.
01 / Industry
Sports Technology
02 / Features
8
03 / Stack
10
04 / Delivery
AI-native
What we delivered on this build
Services
- Backend Development
- API Integration
- Bot Development
- AI / ML Integration
Technology
- Java 17
- Spring Boot 3.3.5
- MySQL
- Hibernate / JPA
- Telegram Bot API
- OpenAI Vision API
- Tesseract OCR
- Google Sheets API
- OkHttp3
- Docker
Project overview from the brief
Original case study content from Ingenious Minds Lab — adapted for this archive.
Bet365 is a Spring Boot backend that monitors multiple Telegram publisher channels and automatically processes parsed signals against the Bet365 platform API. It supports multi-sport, multi-tenant, and multi-account workloads with configurable filters, an account onboarding pipeline, and AI-assisted parsing for screenshot-based messages. The system manages the full request lifecycle — from signal ingestion to execution confirmation — with concurrency controls, fuzzy team name matching, and Google Sheets reporting.
At a glance
- Industry
- Sports Technology
- Capabilities
- 8 features
Challenge & solution how we delivered
01 · Challenge
The core challenge is reliably extracting structured signals from unstructured Telegram messages and images posted by diverse publisher sources, each with its own formatting and conventions. The platform enforces strict session and account-level rules that require careful session management and synchronization to avoid failed executions. Scaling across dozens of channels, multiple sports, and multiple customer accounts simultaneously demands a robust, non-blocking architecture.
02 · Solution
Each publisher source has a dedicated parser for its message format. Image-based payloads are processed via Tesseract OCR with an OpenAI Vision fallback for complex screenshots. A fuzzy name matching engine using Levenshtein distance and cosine similarity resolves team name variations against the platform event catalog. Per-account locks and a concurrent queue manager prevent session conflicts. An onboarding subsystem gradually activates new accounts with controlled low-volume activity across configurable leagues before live signal handling.
What stands out in the product
01 · Highlight
30+ Channel Parsers
Dedicated parser per publisher covering plain text, screenshots, and structured payloads across soccer, basketball, tennis, and more.
02 · Highlight
OCR + OpenAI Fallback
Tesseract OCR handles image payloads; OpenAI Vision fills gaps when OCR output is ambiguous, targeting near-complete signal extraction.
03 · Highlight
Onboarding Pipeline
Automated account conditioning runs controlled activity across configurable leagues and market types before accounts go live.
04 · Highlight
Fuzzy Match Engine
Custom team name resolution using cosine similarity and Levenshtein distance with noise filtering for club suffixes and age-group modifiers.
What the product delivers for end users
01 · Feature
Multi-Channel Signal Parsing
Supports 30+ Telegram publisher channels, each with a dedicated parser that normalizes its unique text and attachment formats into a shared signal schema.
02 · Feature
AI-Assisted Image Parsing
Screenshot-based messages run through Tesseract OCR. When confidence is low, an OpenAI Vision API fallback extracts structured event and market fields from the image.
03 · Feature
Multi-Sport Market Resolution
Resolves markets across Soccer, Basketball, Tennis, Volleyball, Handball, and Ice Hockey — including complex segments such as handicaps, totals, corners, and cards.
04 · Feature
Account Onboarding Pipeline
New platform accounts follow a configurable onboarding flow with gradual, low-volume activity on selected leagues before promotion to live signal processing.
05 · Feature
Fuzzy Team Name Matching
A custom engine using Levenshtein distance and cosine similarity maps publisher team labels (including suffixes like FC, U21, Reserves) to official event names with high accuracy.
06 · Feature
Telegram CLI Management
Operators manage customers, accounts, and channel subscriptions through Telegram commands and interactive menus without direct server access.
07 · Feature
Google Sheets Reporting
Completed actions and system events sync to Google Sheets via the Sheets API for customer-facing dashboards and audit trails.
08 · Feature
Concurrent Multi-Account Execution
Per-account and per-session locks with a queue manager let multiple accounts run in parallel without session conflicts or race conditions.
Product in context visual archive
Turn your roadmap into living AI products.
Tell us what you want to launch — we’ll outline a practical path from first prototype to production, with milestones you can track and no forced long-term contract.
Response time
Under 1 business day
Confidentiality
NDAs on request
First prototype
Often within a week