English · Русский

• I'm a second-year student at Innopolis University.
• I work mostly with Python backend development: FastAPI, SQL, PostgreSQL, APIs, Docker, and testing.
• I also use TypeScript, React, Next.js, Electron, Chrome Extensions, and SwiftUI for web and desktop projects.
• I like projects where I can build the whole thing: API, database, UI, tests, deployment, and documentation.
• Right now I'm learning Go and doing more work with ML / NLP / Computer Vision. I'm open to remote internships, trainee roles, and junior positions.

Backend, data & APIs

Applied ML, NLP & Computer Vision

Web, desktop & browser extensions

Tools, testing & infrastructure

Learning & foundations

An engineering project for the Ozon Tech × Innopolis University selection. I designed a 3D station for measuring item L/W/H on a moving conveyor: compared sensor options, chose LMI Gocator 2880, calculated the layout, and built a synthetic Python PoC that turns point clouds into a WMS-ready result. On 500 Monte Carlo scenarios, 93.6% of accepted measurements were within the required tolerance. The repository clearly separates simulation results from physical tests that still need to be done.

Python · NumPy · SciPy · Open3D · 3D Geometry · Point Clouds · pytest · WMS Integration

A student ML/NLP project for classifying 1,931 labeled Service Desk tickets across 43 categories. I compared TF-IDF, LinearSVC, CatBoost and Qwen3-Embedding 4B/8B pipelines, used grouped validation to reduce data leakage, and built a service that returns TOP-3 categories, recommends a support line and finds similar tickets. Qwen is used as a frozen encoder, while the classification part is trained on the labeled data.

Python · Pandas · scikit-learn · CatBoost · PyTorch · Transformers · Qwen3 Embeddings · FastAPI

A computer vision demo for counting objects that cross a line in video. I use YOLO11n + ByteTrack, keep track IDs, draw the result video, and export crossing events to CSV/JSON. I also tested the pipeline on real videos.

Python · OpenCV · Ultralytics YOLO · ByteTrack · NumPy · Pytest

A full-stack project for tracking a treatment plan and daily intake. I built a FastAPI/PostgreSQL backend and a Next.js/React frontend, with a calendar and diary, private documents and photos, read-only doctor access, Telegram reminders, Excel export, backups, and E2E tests. The app stores the plan entered by the user and does not make medical decisions.

Python · FastAPI · SQLAlchemy · PostgreSQL · Next.js · TypeScript · Docker · Playwright

A full-stack lost-and-found service I worked on. It has a FastAPI/PostgreSQL backend, React frontend, Telegram bot, search and matching, claims, moderation, and Docker Compose deployment.

Python · FastAPI · SQLAlchemy · PostgreSQL · React · Telegram Bot · Docker

A macOS app for organizing Telegram Saved Messages. It syncs messages through MTProto, indexes them locally with SQLite/FTS5, and uses RouterAI for classification. Credentials are stored with Electron safeStorage.

TypeScript · Electron · React · SQLite · Telegram MTProto · AI Integration

A native offline macOS trainer I made for the Russian amateur-radio third-category exam. It has 218 verified questions, spaced repetition, weak-question practice, a 25-question mock exam, a glossary, local progress, and CI/content checks.

Swift · SwiftUI · macOS · Adaptive Learning · Spaced Repetition · Testing · CI

A team project for managing retail orders and inventory. The backend uses FastAPI, SQLAlchemy and PostgreSQL, and the frontend uses React/TypeScript. I worked mainly on the Orders domain: order flows, reservations, and stock changes.

Python · FastAPI · SQLAlchemy · PostgreSQL · React · TypeScript · Vite · Testing

A macOS desktop assistant for live speech and screen questions. Yandex SpeechKit handles STT, questions are detected locally, and several LLMs can answer in parallel. The app is built with Electron, React, TypeScript, Vite, and gRPC.

TypeScript · Electron · React · Vite · gRPC · Speech-to-Text · LLM Streaming · Multimodal AI

A Chrome extension for Yandex Forms and normal web pages. It reads form controls from visible DOM structure, can answer from selected text or screenshots, and has controlled auto-apply, Multi-Check, and Course Pack support.

JavaScript · Chrome Extension · Manifest V3 · Yandex Forms · RouterAI · Semantic DOM · Multimodal AI · Testing

A Moodle-specific QuizMind version. It detects Moodle question types and decides whether to send cleaned HTML, question images, or a screenshot to the model. It also removes review-state clues and old answers before the AI request.

JavaScript · Chrome Extension · Manifest V3 · Moodle · RouterAI · DOM Analysis · Multimodal AI · Multi-Model

A team SaaS-style project with Next.js, NestJS, PostgreSQL, Redis, Prisma, and Docker. It includes authentication, feature flags, remote config, and support/admin workflows. The repository on my profile is a fork because the original development repo was hosted by a teammate.

TypeScript · Next.js · NestJS · Prisma · PostgreSQL · Redis · Docker

• SmartRoute VPN — subscription dashboard with React, Express, Prisma, SQLite, Telegram login, and Docker.
• CRYPTOGRAPHER — Python/PyQt desktop app for encrypting local files and packaging the app for desktop use.

• Getting stronger at Python backend and full-stack development: FastAPI, PostgreSQL, Docker, testing, and deployment.
• Building projects end to end: API, database, frontend, authentication, CI, and deployment.
• Learning more ML / NLP / Computer Vision beyond just using ready-made models.
• Learning Go and looking for a remote internship, trainee role, or junior position where I can work on real tasks and learn from a team.

I'm open to remote internships, trainee roles, and junior developer positions, mostly around backend, full-stack development, automation, desktop tools, and Applied ML / NLP / Computer Vision.