Grigorij

TinderGPT is designed as an AI assistant to streamline the dating experience, sparing users from the time-consuming and often uncertain process of messaging on dating apps. This AI handles everything on behalf of the user—from initiating conversations with messages tailored to the profile in question, crafting engaging interactions, to arranging dates and providing the user with the match’s contact information.

As CTO at PrimerAI, I developed an intelligent platform for personalized cosmetics suggestions using ML and AI, including OpenCV, GPT-4, and K-means clustering for skin tone detection and categorization. I employed a holistic approach in developing and implementing this web app, which analyzed users' skin tones using computer vision and unsupervised ML algorithms. My contributions covered all aspects of the project, including designing and building the frontend and backend of the web app, as well as administering Linux servers. This ensured a seamless, end-to-end product experience that catered to our users' needs.

In this project, I focused on implementing small open-source models, such as Zephir-7b, on single-board computers like the Raspberry Pi. This setup enabled the creation of a robot controlled by an LLM. Working with such compact models presents numerous challenges; one significant issue is the reduced level of intelligence compared to larger models like GPT-4. To address this, the project required fine-tuning the open-source LLM to achieve satisfactory results.

At WildFoundry, I held the role of developer advocate, producing educational programming content through articles and YouTube videos while offering client support. A prominent GPT use case from my experience involved analyzing the apps that users ran on their devices. To overcome the challenge of manually examining tens of thousands of users, I developed a script that scraped user pages one by one and analyzed the content via the GPT API before storing the results in a pandas dataframe. This approach enabled an unbiased and comprehensive analysis of user applications, yielding accurate numbers and insights that outperformed traditional survey methods. The ensuing report allowed the identification of target customer groups and guided the company in shaping its customer relation strategy.

During the development of a web application for the TakŻyli startup (takzyli.pl), I employed AI agents to automate the code writing process. The project utilised Microsoft Autogen and CrewAI as multi-agent collaboration orchestrators to streamline the entire application development process. These agents were equipped with tools for checking existing code, modifying it, and inserting new code. They also had the capability to review application logs after implementing changes to self-correct and to inspect the frontend's appearance. The use of multi-agent automation in this project demonstrated its vast potential in enhancing development efficiency and innovation.

Dating apps, while popular, had users spending excessive hours daily, leading to notable frustration. Research from smalltalks.ai indicated that the average daily usage was a substantial 1.5 hours, with a staggering 70% of users experiencing anxiety post-use. To address this, I developed an application powered by the advanced GPT-4 artificial intelligence. This AI not only mimics human interaction but crafts initial messages based on individual profiles, engages in meaningful conversation, and establishes emotional connections using a meticulously curated pick-up database. Once a rapport is established, it proposes a real-life meeting and notifies the human user, thus streamlining the dating process while reducing user stress.

GregTraider is a streamlined framework designed for automated online trading and trading strategy backtesting using historical data. It simplifies the process by allowing users to write a single strategy for both backtesting and online trading, eliminating the need to rewrite and adjust strategies for different platforms. The primary benefit of using Gregtraider is that you only need to write and adjust your trading strategy once for backtesting, and it is immediately ready for real trading. This saves you the tedious work of rewriting and testing your strategy multiple times.