Designing and implementing automated pipelines for integrating LLMs into user workflows. Working with large language models via OpenAI, Claude, Gemini, Perplexity APIs. Using OpenRouter for multi-model routing and cost/performance optimization. Fine-tuning inference parameters (Top-P, Top-K, Temperature, Reasoning Effort) for task-specific behaviour. Creating prompts from scratch and refactoring existing user and developer prompts. Applying prompt design patterns (persona, audience, template, recipe, visualisation generator). Using advanced prompting techniques: zero-shot, few-shot, chain-of-thought, self-consistency, tree of thoughts, directional stimulus prompting, reflection. Writing prompts for multiple domains and use cases with different constraints and objectives. Testing and benchmarking prompts using built-in and custom evaluators. Designing Retrieval-Augmented Generation (RAG) pipelines. Working with vector databases for semantic search and contextual retrieval. Optimising chunking, embedding strategies, and retrieval parameters. Developing a marketing AI agent that retrieves LinkedIn user data via API and generates personalised, targeted advertising content. Building an internal MVP consisting of multiple agents for:
- prompt generation and improvement
- input/output analysis
- basic automated testing of LLM outputs
Creating a personal assistant for:
- monitoring and notifying about relevant job openings based on user profile
- providing on-demand market data (stocks, currencies)
Deploying AI services in Google Cloud. Containerising applications and managing environments with dependencies. Automating workflows and integrations using n8n.

Developing models for detecting DeepFake content (photos, videos, audio) Researching model architectures, working with pre-trained transformers. Creating a complete pipeline for generating and recognising DeepFake data. Deploying and configuring the FaceFusion complex for local training data synthesisBuilding datasets and automating collection. Writing scripts for automated data collection, filtering, and annotation. Quality control of datasets, integration of results into the training pipeline. Quality validation of sets using metrics and manual verification Local deployment of models and computing infrastructure. Deployment of ML models on local servers with GPUs. Containerisation of models, building environments with dependenciesInfrastructure and clouds (Azure). Configuring network infrastructure in Microsoft AzureTesting and optimising performance for computational tasks. R&D / research and experiments. Searching for and testing new approaches to detecting fraud. Conducting comparative analysis of models, tuning hyperparameters. Documenting results and preparing internal reports

Developing and automating internal services (Java-Script, GCP). Creating and optimization main part of internal CRM system. Creating automatically informing service for clients. Creating a simple internal system for accounting.

Development of microservice architecture (Java, Spring). Participation in the creation of microservices: authentication server, SMTP serverImplementing Security Logic Using JWT. Creating templates via Thymeleaf, authorisation logic and interface. Testing and code quality. Code coverage with unit tests (JUnit, Mockito). Refactoring and maintaining service stability in production. Infrastructure and containerisation. Deploying services in Docker, configuring environments. Support and update microservice configurations. Working with databases using MongoDB (NoSQL) to store sessions and logsWorking with PostgreSQL on EC2 instances. Scheme structuring and query optimisation. Cloud infrastructure (AWS). Configuration and integration of EC2, S3, Lambda, Amplify, API Gateway. Working with DynamoDB as a NoSQL solution for individual services. Building cloud architecture for web applications. Team interaction and deployment support. Participation in release planning, staging and production deployment. Documenting configurations, updating service interaction diagrams.