Anton K.

I am a Machine Learning Engineer and AI Automation Developer experienced in building advanced AI systems, RAG applications, multi-agent architectures, and Computer Vision solutions using Python, PyTorch, Transformers, OpenAI API, LangGraph, FastAPI, and vector databases. I work on full-cycle machine learning development and LLM integration: engineering autonomous AI agents, evaluating and optimizing LLM pipelines, implementing multimodal systems, improving model inference performance, and deploying robust AI architectures to production environments. Please check below more details about some of my projects. LLM Evaluation Framework for Enterprise AI Platform - as an ML Engineer I worked on a comprehensive evaluation pipeline for a multi-agent system to ensure factual grounding and mitigate deployment risks. Contributed to synthetic data generation, live system validation for vector indexing, and adversarial testing against prompt injections and hallucination provocations. Used Python, LLM-as-a-judge, DeepEval, RAG, OpenAI API. GenAI-Powered Supply Chain Digital Twin - as an ML Engineer I worked on an AI-assisted ETL pipeline and conversational agent for mapping complex data schemas. Contributed to LangGraph-based multi-agent orchestration, secure tool integration via Model Context Protocol (MCP), and the development of an interactive data sandbox for text-to-SQL queries. Used Python, LangGraph, OpenAI API, Model Context Protocol (MCP), Snowflake, Neo4j, Redis, FastAPI, Prefect. AI-Powered Mobile Automation Agent - as an AI Automation Developer and Backend Engineer I worked on an autonomous mobile agent executing complex tasks on Android devices using multi-modal LLMs and computer vision. Contributed to the core ADB automation pipeline, serverless GPU model deployment, knowledge graph integration for UI states, and human-in-the-loop training loops. Used Python, Transformers, OpenAI API, FastAPI, Neo4j, PostgreSQL, AWS, Docker, RunPod. Generative AI Text-to-Video Avatar Synthesis - as an ML Engineer I worked on a pipeline transforming static images and text into dynamic speaking avatars. Contributed to custom Text-to-Speech engineering, one-shot avatar animation, and cross-modal phoneme-to-viseme synchronization. Used PyTorch, Hugging Face, Transformers, Wav2Lip, OpenCV, FFmpeg.