Staff AI & Machine Learning Engineer (Tech Lead) at TaxDome
Serbia
Dmitriy Ulybin
Council Member
Dmitriy Ulybin is a Staff AI and Machine Learning Engineer serving as Tech Lead at TaxDome, with over seven years of experience building AI products, machine learning systems, and engineering teams from the ground up. His work spans the full arc of production AI development, from defining the problem and standing up infrastructure to shipping features and scaling the teams that maintain them.
At TaxDome (April 2025 to present), a practice management platform serving more than 10,000 accounting firms, Dmitriy Ulybin joined as the company's first ML hire with no existing infrastructure, no training data, and no team. Within six months he launched TaxDome's first AI product into production, and within eleven months he had delivered three AI products, growing the function from a solo role to a four-person team spanning ML engineering and systems analysis. He designed a hybrid pipeline that combines large language models, computer vision, business rules, and Agentic RAG, achieving approximately 0.96 F1 and 99% system reliability. The AI-generated checklist feature he shipped was adopted by more than 2,000 accounting firms, reducing manual processing time by 83% and saving an estimated 30,000 hours across a single tax season. He subsequently expanded the product portfolio to include Document Tagging, Client-Document Matching, and AI-powered Tax Return Summaries.
Prior to TaxDome, Dmitriy Ulybin spent over two years at MTS Group (December 2022 to March 2025), a telecom and digital products company serving more than 80 million subscribers, where he held the role of Senior Machine Learning Engineer. He architected and shipped a RAG-based search and recommendation system that served millions of users and improved retention by approximately 20% through intent classification, semantic retrieval, and named entity recognition. He also built an ML-powered vulnerability detection pipeline integrated directly into CI/CD workflows, cutting false positives by 70% and reducing manual review burden on application security teams. During this period he mentored two junior ML engineers to mid-level, established production ML and MLOps practices across experiment tracking, inference optimization, monitoring, and containerized microservices, and shipped four production ML services in two years.
At TaxDome (April 2025 to present), a practice management platform serving more than 10,000 accounting firms, Dmitriy Ulybin joined as the company's first ML hire with no existing infrastructure, no training data, and no team. Within six months he launched TaxDome's first AI product into production, and within eleven months he had delivered three AI products, growing the function from a solo role to a four-person team spanning ML engineering and systems analysis. He designed a hybrid pipeline that combines large language models, computer vision, business rules, and Agentic RAG, achieving approximately 0.96 F1 and 99% system reliability. The AI-generated checklist feature he shipped was adopted by more than 2,000 accounting firms, reducing manual processing time by 83% and saving an estimated 30,000 hours across a single tax season. He subsequently expanded the product portfolio to include Document Tagging, Client-Document Matching, and AI-powered Tax Return Summaries.
Prior to TaxDome, Dmitriy Ulybin spent over two years at MTS Group (December 2022 to March 2025), a telecom and digital products company serving more than 80 million subscribers, where he held the role of Senior Machine Learning Engineer. He architected and shipped a RAG-based search and recommendation system that served millions of users and improved retention by approximately 20% through intent classification, semantic retrieval, and named entity recognition. He also built an ML-powered vulnerability detection pipeline integrated directly into CI/CD workflows, cutting false positives by 70% and reducing manual review burden on application security teams. During this period he mentored two junior ML engineers to mid-level, established production ML and MLOps practices across experiment tracking, inference optimization, monitoring, and containerized microservices, and shipped four production ML services in two years.
Before that, Dmitriy Ulybin worked as a Machine Learning Engineer at Xena Exchange (January 2021 to November 2022), a fintech platform focused on crypto market intelligence and trading signals. He built a news intelligence pipeline covering ingestion, ETL, named entity recognition, sentiment scoring, and automated summarization to support real-time trading signals for institutional clients. He designed uplift models for marketing campaigns that increased user conversion by 22%, and improved signal accuracy by 30% through fine-tuned sentiment analysis and entity linking for cryptocurrency token impact classification.
Earlier in his career, Dmitriy Ulybin served as a Data Scientist and ML Engineer at EPAM Systems (February 2019 to December 2020), a global software engineering and consulting company, where he delivered end-to-end machine learning solutions for enterprise clients across the oil and gas, transportation, and aviation sectors. His work there included predictive maintenance and forecasting solutions that reduced unplanned downtime by approximately 15%, as well as ranking and ML pipeline solutions that improved click-through rate by roughly 12% and reduced iteration cycles by around 40% through automated workflows.
Outside his primary roles, Dmitriy Ulybin has competed in ML challenges, placing Top 5 out of 80 teams in the Agentic RAG Legal Challenge at Dubai AI Week (Machines Can See 2026), and achieving a public leaderboard score of 0.8536 in Data Fusion 2026 Track 2, for which he received the Companion nomination for releasing the best open-source solution, publishing his complete seven-script pipeline for other competitors to build upon.
Dmitriy Ulybin holds a degree in Big Data and Machine Learning from ITMO University (2022). His technical expertise spans artificial intelligence, machine learning, large language models, retrieval-augmented generation, agentic workflows, document AI, natural language processing, MLOps, inference optimization, Python, and AWS.
Earlier in his career, Dmitriy Ulybin served as a Data Scientist and ML Engineer at EPAM Systems (February 2019 to December 2020), a global software engineering and consulting company, where he delivered end-to-end machine learning solutions for enterprise clients across the oil and gas, transportation, and aviation sectors. His work there included predictive maintenance and forecasting solutions that reduced unplanned downtime by approximately 15%, as well as ranking and ML pipeline solutions that improved click-through rate by roughly 12% and reduced iteration cycles by around 40% through automated workflows.
Outside his primary roles, Dmitriy Ulybin has competed in ML challenges, placing Top 5 out of 80 teams in the Agentic RAG Legal Challenge at Dubai AI Week (Machines Can See 2026), and achieving a public leaderboard score of 0.8536 in Data Fusion 2026 Track 2, for which he received the Companion nomination for releasing the best open-source solution, publishing his complete seven-script pipeline for other competitors to build upon.
Dmitriy Ulybin holds a degree in Big Data and Machine Learning from ITMO University (2022). His technical expertise spans artificial intelligence, machine learning, large language models, retrieval-augmented generation, agentic workflows, document AI, natural language processing, MLOps, inference optimization, Python, and AWS.