Danil Baranczov

I study at two universities: Artificial Intelligence at Central University and Applied Mathematics and Informatics at HSE University's Faculty of Computer Science. I have 2 years of commercial experience. I am a winner of the regional stage of the All-Russian Olympiad for School Students in Mathematics and the All-Siberian Open Olympiad for School Students, and a prize winner of T-Bank's Datathon 2.0. I think analytically, learn quickly, pay close attention to detail, and see tasks through to completion.

Work experience

Machine Learning Engineer

March 2026 - October 2026

Technology stack upgrade

Upgraded the MLOps framework from 0.1.x to 0.2.x and migrated GitLab CI to pipeline v2. Moved the project to Python 3.11 and verified compatibility with Python 3.12. Upgraded the train, QA and prod Airflow instances to 2.10.5, migrated models from S3 to Model Registry, and adapted the loading code and configurations. Verified training and QA DAGs, as well as inference. Retrospective testing confirmed that model scores remained unchanged after the migration.

172T model development

Prepared a local dataset for retraining: corrected SQL scripts and Airflow DAG logic, replacing the use of two bases with a single one. Studied the model architecture and previous experiment results, then formulated 10 hypotheses for improving quality. Together with the team, tested the hypotheses on the prepared data, compared experimental results and identified changes that improved the model’s metrics.

ML System Doc update

Reviewed the existing documentation and code for every model in the production repository. Compared model descriptions with their implementations and collected questions about data preparation, training and inference. Met with model developers to clarify the details and used their answers to update ML System Doc, bringing the documentation into line with the current implementation.

Frontend Developer

February 2026 - July 2026

Feedback service

Developed interfaces for creating, viewing and processing user requests. Implemented validated forms, request lists with search, filtering and sorting, and detailed cards showing status and processing history. Designed interface states for empty results, data loading and errors.

Analytics service

Built analytics dashboards with key metrics, charts and tables. Implemented filtering by period and category, switching between data views, and viewing detailed information for selected metrics. Developed reusable components for consistent presentation of analytics.

API integration

Connected interfaces to backend services: implemented data retrieval and submission, response handling, and content updates after user actions. Added loading states, error messages and notifications about operation results. Coordinated data formats and interface behavior with backend developers.

Machine Learning Engineer

November 2025 - March 2026

ML pipeline development

Designed the architecture of the intelligent core for an on-call support incident analysis system based on a hybrid ML pipeline. Combined speech recognition, transcript processing and report generation into a sequential workflow for processing calls.

Speech recognition

Integrated Vosk for local, real-time transcription of call audio streams. Implemented low-latency CPU processing so that the conversation text entered the system as the incident was being discussed.

Report generation

Implemented a use-case layer to buffer transcripts and group them by session. Connected the Groq API with the Llama-3 70B model to generate reports from call transcripts at more than 200 tokens per second.

01/01

Projects

Analyzed and prepared more than 500,000 observations, accounting for temporal data shift. Built a CatBoost and LightGBM ensemble with optimized weights, outperforming the baseline and achieving a ROC-AUC of 0.74.

GitHub Kaggle

Analyzed data from 125,000 customers and built a CatBoost model with a ROC-AUC of 0.9012. Used SHAP to identify key churn drivers, defined five risk segments and prepared recommendations for customer retention.

GitHub

Developed a PyTorch neural network solution to predict the number of positive and negative reviews. Combined numerical features with game descriptions, achieved an MAE of around 180 and placed 21st out of 242 competition participants.

GitHub Kaggle

Developed a web service for viewing weather forecasts along a route on a map. Implemented a client-server architecture, an external API integration, request handling and weather data visualization.

GitHub

As part of a team, I am developing a personal assistant for managing tasks, finances, health and wardrobe. The project includes a React and TypeScript interface and a trend discovery service built with Node.js and Fastify, featuring web search, SQLite memory and LLM integration through OpenRouter or Ollama.

GitHub

Developed a Telegram Mini App for ticket sales and student event management. Implemented electronic ticket issuance, an auction, promo codes, a merchandise shop and a token system with QR codes. Built a React and TypeScript interface, a FastAPI backend with PostgreSQL, and integrations with Telegram and Google Sheets.

GitHub

01/01

Technology stack

Classic ML

Python, NumPy, Pandas, scikit-learn, CatBoost, LightGBM, Optuna, data analysis, data preprocessing, feature engineering, model ensembles, cross-validation, SHAP.

MLOps

Docker, Git, GitLab CI/CD, Apache Airflow, S3, Model Registry, Linux, training and inference automation, ML pipeline testing, model monitoring.

Frontend

HTML, CSS, JavaScript, TypeScript, React, responsive layouts, REST API, data visualization, npm, Vite.

Backend

Python, Flask, FastAPI, SQL, PostgreSQL, REST API, client-server architecture, external service integration, testing, Docker.

Deep Learning

Python, PyTorch, neural networks, NLP, embeddings, transformers, LLMs, multimodal models, GPU training, speech recognition.

Mathematics

linear algebra, calculus, probability theory, mathematical statistics, discrete mathematics, optimization methods, hypothesis testing, A/B testing.

MY CONTACTSMY CONTACTSMY CONTACTSMY CONTACTSMY CONTACTSMY CONTACTSMY CONTACTSMY CONTACTS

Danil Baranczov

Work experience

Projects

Technology stack

Classic ML

MLOps

Frontend

Backend

Deep Learning

Mathematics

Contacts