Yegor — ML Specialist
Languages
Frameworks
Tools
Databases
Yegor built a full-scale engine for evaluating the financial stability and reliability of corporate counterparties, replacing a previously manual and error-prone workflow. He designed unified data ingestion pipelines that integrated large volumes of heterogeneous information from public registries, financial disclosures, and internal systems, applying rigorous cleaning and deduplication to ensure high-quality inputs. Through deep exploratory analysis, he identified the key drivers of counterparty risk, developed a structured risk-factor framework, and engineered domain-specific reliability metrics. He then built and optimized classification and regression models that forecast financial distress, improving predictive accuracy by 20%. The resulting system provided analysts with clearer, more interpretable signals, reduced manual verification, and significantly strengthened overall risk-assessment capabilities.
Read moreYegor architected an enterprise Retrieval-Augmented Generation platform that transformed internal knowledge access across the organization. He built a high-fidelity semantic retrieval system capable of extracting nuanced policy, compliance, and operational information from large volumes of internal documentation, enabling instant, context-aware search. He integrated LLMs to serve as a knowledge assistant that provided cited, verifiable answers, sharply reducing turnaround time for routine employee inquiries. Yegor also customized and domain-adapted language models to align with internal terminology, workflows, and compliance requirements, significantly improving accuracy and relevance. The platform cut document search time from minutes to seconds, reduced repetitive requests to support teams, and improved decision-making by delivering precise, traceable information on demand.
Read moreAt TEC, Yegor worked on a cross-functional ML and Document Intelligence team developing a large-scale multilingual OCR and academic evaluation platform. He contributed to building OCR and layout-understanding pipelines capable of interpreting complex academic documents, including tables, seals, and stamped forms, and supported automation of entity extraction and evaluation parsing. He participated in creating models for course-to-course matching, GPA computation, credit equivalency, academic level classification, and international grade normalization. Yegor also helped implement domain-specific logic for extracting clinical hours in nursing programs and supporting engineering and teaching tracks to meet accreditation and compliance standards.
Read moreAt Scrivas, a healthcare AI platform for automated clinical documentation, Yegor served as the sole ML engineer responsible for designing, deploying, and scaling all core AI systems. He built a multilingual speech recognition pipeline with diarization and word-level bilingualism detection, integrated LLM-driven clinical note generation tools for SOAP, HPI, ROS, PE, and A&P templates, and developed clinical decision-support logic using ICD-10, HCC, and E&M guidelines. He implemented PHI masking and prompt auditing to ensure security and compliance, and designed the full production architecture for writing notes to EHRs through HL7 and FHIR, including retries, fallbacks, and SLA monitoring. His work spanned ASR, LLM engineering, healthcare compliance, and production MLOps.
Read more
Second Talent data shows AI-assisted coding can save 30% to 75% of a developer's time. Now, those gains aren’t the result of faster code suggestions. They come from AI systems...
Elite Brains data shows that as of 2025, roughly 41% of code worldwide is now AI-generated. This is clear proof of a paradigm change: development workflows are now shifting from...
