Ruslan Belkov

Ruslan Belkov

Forward Deployed Software Engineer · AI · Full-stack

Full-stack software engineer focused on AI, with 6+ years of experience. Deep expertise in Python, TypeScript, and software architecture, backed by an academic background in Data Science & AI, Cybersecurity, and Robotics.

My experience spans web development, production AI systems, industrial computer vision, self-hosted AI deployment, AI-powered cybersecurity, open source, and entrepreneurship.

I combine a strong engineering foundation with effective use of AI-native development workflows, leveraging coding agents, custom agent skills, prompt design, diverse model integrations to accelerate delivery while maintaining code and product quality.

I could take ownership of the entire product lifecycle: from asking the right questions and clarifying customer needs to design, architecture, implementation, documentation, and DevOps. Comfortable working directly with customers, navigating ambiguity, and shipping in fast-paced environments.

Co-founder of a student developer community whose services are used by over 90% of Innopolis University students. Experienced in rapid prototyping, scalable architecture, and competitive programming. Captain of the Tatarstan National Team in Product Programming and holder of the Master of Sports title in programming. An enthusiastic technophile with a passion for innovation, I enjoy tackling complex challenges, exploring emerging technologies, and experimenting with unconventional setups, including Nobara OS, the QGMLWY keyboard layout, and vertical mice.

Education

Innopolis University — B.Sc. in Data Science & Artificial Intelligence

Bachelor's Degree, Russia – 2022-2026, the entire program was taught in English

Exchange Semester at Skoltech, Moscow, Russia – Spring 2025

Winter Artificial Intelligence Program at Beijing Institute of Technology, Zhuhai, China – January 2026, was awarded for the outstanding performance with project related to Embedded AI

Innopolis University — M.Sc. in Robotics

Master's Degree, Russia – 2026-present

Professional Experience

Software Engineer at R77 AI, Contract

Jul 2024 - Present

Developed a Retrieval-Augmented Generation (RAG) system for AlfaBank to assist third-party developers in integrating with the bank’s APIs (developers.alfabank.ru). The assistant was embedded directly on the website in a chat format, providing real-time guidance and reducing support load. Achieved a 70% exact-answer rate and decreased the average number of daily emails to the RAN team by 30%.

Developed a multilingual AI chatbot for a Bureau of Kazakhstan to assist users with statistical forms, information searches, and technical support. The bot processes queries in real time, collects user feedback, and integrates with the contact center's CRM to handle unresolved issues and improve service quality.

Developed a Telegram bot for restaurants to automate feedback collection, analysis, and reporting from platforms like Google Maps, Yandex, 2GIS, Instagram, and TripAdvisor. Integrated ChatGPT for sentiment analysis and actionable insights, provided real-time notifications, daily reports, and enabled staff feedback submission.

Developed an AI-powered Telegram bot for a Ministry of Kazakhstan, enabling users to ask questions, receive instant responses, and provide feedback. The bot collects consent for data processing, supports follow-up queries, and escalates unresolved issues to human operators.

Developed a Telegram bot for a loyalty program in a store specializing in candles. The bot generated PDF files with QR codes formatted for adhesive label printing, which were applied to products. Customers scanned the QR codes to access the bot, earn rewards, and progress through the loyalty program tiers, streamlining the process of engagement and rewards tracking.

Software Engineer at Innopolis University, Full-time

Oct 2024 - Jun 2026

Working as a Software Engineer at Advanced Development Department of Innopolis University developing backend services and APIs in Python and FastAPI for an industrial direct metal deposition system. My responsibilities include building AI and computer vision pipelines for industrial applications, developing algorithms for 3D mesh, STL, and point-cloud processing, implementing slicing and surface deviation analysis, and contributing to robot-controller communication, trajectory planning, and inverse kinematics for a Kawasaki RS080N-based robotic setup.

Full-stack Software Engineer at Trippy, Contract

Summer 2026

Developed a web platform and AI-powered automation workflows for visa application services, including document preparation, monitoring and booking appointment slots at visa application centers, arranging flight and hotel reservations, and processing travel insurance requests.

Startups & Leadership

one-zero-eight

Sep 2022 - Present

I am a co-founder of one-zero-eight, a student community at Innopolis University (IU). Our mission is to create the perfect environment for student life. We develop projects, participate in competitions, host events, and produce podcasts. Over 90% of students use our services. I lead a team of 20+ members, conduct interviews for new recruits, mentor them, also I design and architect solutions, and contribute by building bots and backend systems. Carried out external communications with various departments.

“PromOblako” digital platform

Dec 2025 - Present

Co-creating PromOblako — a production monitoring platform with a long-term vision of digital twins for industrial facilities. We build edge devices that count products on conveyor lines, send telemetry to the cloud, and visualize production metrics as charts and tables. My contributions include product architecture, backend and cloud infrastructure, and data processing pipelines.

ООО "РАСПИСАНИЕ БЕЗ ХЛОПОТ"

Sep 2024 - Jan 2026

I received a "Student Startup" grant of 1 million rubles to develop a startup focused on building a technological ecosystem for aggregating, managing, and visualizing schedules and events in universities and schools. Currently, the project is in the stage of establishing its legal framework and drafting a business plan. We have also made significant progress in automating scheduling processes in educational institutions to enhance efficiency and improve the user experience for students and staff.

Overlord Systems

Nov 2023 - Nov 2025

I am a co-creator of a mobile information system for golf clubs. It helps golf course operators manage their fleet of golf carts and enhances the playing experience for golfers. The system is currently successfully operating in golf clubs across Russia. My contributions include backend architecture and development, as well as mobile application implementation. Additionally, I have actively participated in negotiations with golf clubs and delivered product demonstrations to potential partners and customers.

Open-Source Projects

pdf-extraction-agenda

Curated research repository comparing PDF/OCR-to-Markdown extraction pipelines, including open-source tools, proprietary APIs, specialized VLMs, and general multimodal models. Aggregates benchmark results, capabilities, licenses, API availability, and deployment trade-offs to support technical evaluation of document processing stacks.

meow-embed

FastAPI-based embedding server and Python client for retrieval systems. Provides dense, sparse, and BGE-M3 embeddings in a single request with efficient transport via base64-encoded vectors and gzip-compressed requests/responses. Also supports client-side caching and integration with rerankers/CrossEncoders.

lee-label

Local-first, cross-platform desktop application for labeling computer vision datasets without accounts, cloud uploads, or external storage. Built with Electron, SolidJS, and SQLite, it provides fast image and mask rendering through preloading and WebGL shaders, streamlined keyboard-driven workflows, and AI-assisted annotation using SAM-HQ, SAM 3, and SAM 2.1.

fastapi-template

CookieCutter-based starter for building FastAPI applications with uv. It supports PostgreSQL with SQLAlchemy or MongoDB with Beanie for databases, integrates Ruff for formatting and linting, and enables deployment with Docker and GitHub Actions. It also includes TUI for adding new routers and generating boilerplate code.

fastapi-derive-responses

Python library that extends FastAPI by automatically generating OpenAPI responses based on your endpoint's source code. This plugin helps you reduce boilerplate code by deriving response statuses and descriptions directly from your code.

fastapi-swagger

Plugin that patches the FastAPI app to include a latest Swagger UI distribution. Also, it works locally and does not depend on the @tiangolo domains and CDNs. It updates automatically on each Swagger UI release.

fastapi-mock

A utility for FastAPI that simplifies the creation of mock endpoints during development. It allows returning generated examples instead of HTTP 500. Examples are based on response models. Advanced features include JSON schema examples, field examples and defaults, and support for custom example providers such as Faker.

Hackathons

Russian Championship in the discipline “Robot Programming”

2025 Winner

Worked on programming an educational robotic manipulator Promobot mEDU: implemented motion programs and task logic, tuned trajectories and movements, and validated execution on the real hardware during the competition.

Tenderhack Kazan

2025 Winner

Built a product grouping / aggregation tool for a supplier portal: automatic aggregation of similar SKUs (СТЕ), manual group management, and sorting by key attributes to make product organization faster and clearer. Implemented ML + vector search approach: embeddings, clustering with AgglomerativeClustering, metrics learning (ITML), semantic search.

  • FastAPI
  • PostgreSQL
  • Scikit-learn
  • Qdrant
  • Docker Compose

Rosneft Student Hackathon

2025, 4th place

Developed a computer vision model to automatically segment map regions with different oil-well placement patterns (48-hour online hackathon). I was disqualified because of ultralytics licensing issues, so now I am wondering about my own MIT library for YOLO models.

  • ultralytics YOLO
  • Sliced Aided Hyper Inference

Archipelago 2025: Processing datasets from drones

2025 Prize Winner

Developing computer vision pipeline with YOLOv8 for detecting lost people on large UAV images. I also searched for additional dataset from open and closed sources, and collected the largest open source dataset on this problem. I was awarded by the RAIS of the Republic of Tatarstan for my contribution to the development of unmanned vehicles.

  • ultralytics
  • s3
  • numpy
  • Pillow
  • scikit-learn
  • httpx

Russian Championship in the discipline “Product Programming”

2024 Winner

Backend development of a search and viewing system for sports events based on the Unified Calendar Plan of the Russian Ministry of Sports, parsing of pdfs (2 days).

Backend development of a platform designed for managing sports competitions in the field of competitive programming and providing informational support to sports federations (2 days).

  • FastAPI
  • MongoDB
  • Beanie
  • pandas
  • pdfplumber
  • pypdf
  • ICalendar

Tatarstan Championship in the discipline “Product Programming”

2024 Winner

Backend development of a platform for conducting surveys targeting respondents in the HR field (2 days).

  • FastAPI
  • MongoDB
  • Beanie

Frontend & Backend development of a browser-based Mahjong game with a car thematics (2 days).

  • React
  • Vite
  • shadcn
  • FastAPI
  • MongoDB
  • Beanie

INNOGLOBALHACK: MTS: AIDocProcessing track

2024 Winner

Development of an intelligent RAG system based on scans of financial documents with interactive charts (3 days).

  • RAG Llama 3.2 & Chroma
  • Open WebUI
  • TableTransformer

Technological competition “Autonomous search. Soft"

2024 Finalist

Development of a remote computer vision pipeline for the task of Search & Rescue (70 days)

  • YOLO ultarlytics
  • SAHI
  • TensorRT
  • NVIDIA Jetson

E-CUP: Everything as Code

2024 Participant

Development of an ML model that will determine whether products are duplicates based on names, attributes and images (30 days).

  • polars
  • catboost

Russian Championship in the discipline “Product Programming”

2023 Finalist

Backend development of system for monitoring PostgreSQL Database in Telegram Bot (2 days).

Backend development of a digital solution for testing users on basic cybersecurity knowledge (2 days).

  • FastAPI
  • PostgreSQL
  • SQLAlchemy
  • Grafana
  • Prometheus

ISS Hack: Firearms detection

2022 Participant

Development of a firearms detector based on CCTV Cameras.

  • Python
  • Torch
  • OpenCV