Our cases
Task: To implement an API for the Telegram web-app application.
Clients came with a description of the functionality of an application similar to the one used for airdrop Notcoin. The app allows users to register, purchase a “pass” and accumulate points by “tapping” a button on the smartphone screen. Within the app, it was required to implement multiple levels of users, depending on which level rewards are awarded. It was also necessary to customize the saving of user progress and the system of charging referral rewards.
Solution: A set of API-endpoints was implemented, which can be accessed by the web application when users perform actions inside the game. Users can use referral links for registration and in this case the system automatically embeds the user into an MLM system with a specified maximum number of direct followers.
User progress and game sessions are stored in the database.
Solution features:
A modern FastAPI framework is used to implement the API.
PostgreSQL is used as a database
The application and database are packaged in Docker containers for easy deployment of the solution on production servers and further scaling.
The search for a new user inclusion point in the MLM system is performed at the database level using the Breadth-First Search (BFS) algorithm. This allows shifting part of the load from the API to the database server and optimally utilizing the client’s IT infrastructure.
Programming languages – Python 3.11, SQL
Task: To implement a system that allows to manage a large number of telegram accounts and automatically perform tasks in the Hamster Kombat web-app.
Solution: Developed a set of scripts that allows to automatically launch the application on behalf of any number of registered Telegram users and perform in-game actions to fulfill the conditions for receiving air-drop. To reduce the probability of detection and blocking of the system by Hamster Kombat developers, the use of individual proxy accounts is implemented. A monitoring system is implemented to control performance.
Solution features:
To reduce the risk of blocking, each user added to the system is assigned information about the mobile device they use. This information is transmitted to the application in the headers of network requests and thus each user from the point of view of the application uses his own, separate mobile device. Combined with the use of individual proxies, this provided a sufficient level of system masking – no account was banned.
User data, information about their devices, and proxy parameters are stored in PostgreSQL in encrypted form. This has ensured a sufficient level of security and reliability of data storage
The system works in multi-threaded mode, which allows you to run a large number of processes simultaneously and manage hundreds of users on one server
The system has an internal API for building a microservice architecture.
Microservice architecture allows increasing the number of managed accounts both by increasing the server capacity and by using several servers.
InfluxDB database and dashboards in Grafana are used to monitor system performance and progress of each account.
The modular implementation allows users to be connected to new applications in the future. Which was later used to increase the profitability of the system.
Task: to automate the selection of optimal trading strategy parameters.
Clients came with a ready-made trading algorithm described in pine-script. The developer was required to implement a solution that would allow to search hundreds of thousands of combinations of algorithm parameters within reasonable time and check the result of the algorithm with these parameters on historical data.
Solution:
A set of scripts was developed that:
automatically downloads historical data using API of Binance and Bybit exchanges.
Selects a set of parameters that provides maximum profit at a given maximum capital drawdown on a given set of historical data. (For example – search for a combination that provides maximum profit at a drawdown not exceeding 15% on data from 01.01.2020 to 01.01.2024).
Forward testing is used to verify the selection quality. That is, after the next set is obtained that satisfies the specified profit/drawdown indicators, the parameters are tested on a new date range in order to reduce the effect of “retraining” of the algorithm. And if the forward testing also shows satisfactory results, the parameter set is saved to the results list.
It saves the obtained results to a text file as a list of parameter values. It also saves the code of the pine script strategy with the best selected parameters into a text file. The code can be copied into TradingView and run.
Solution Features:
A genetic algorithm is used to find the optimal set of parameters, which finds a solution much faster than a direct enumeration of all possible combinations of parameters.
For additional acceleration we implemented parallelization of calculations on available CPU cores.
The calculations are performed using the Numba library, which reduced the computation time by 1000 times compared to traditional methods (pandas, numpy) adopted in python for working with data.
The solution is packaged in a Docker container, which allows you to quickly deploy it on any modern server or personal computer.
The solution allows you to simultaneously (in parallel) perform calculations for any number of trading pairs.
Calculation parameters – list of trading pairs, maximum allowable drawdown, minimum profit level, back-testing range and forward-testing range – are set in a file. After that you can run the script, which will perform all further work autonomously. That is, the user can run the script at night or on weekends and return to work when all the results are calculated.
The programming language is Python 3.11
It is possible to extend the solutions to search parameters for other popular crypto exchanges.
