Ivan Smirnov

Machine Learning Engineer

Munich, Germany

I'm a Machine Learning Engineer specializing in AI systems at scale. With a B.Sc. in Artificial Intelligence from THD, I focus on NLP, Retrieval-Augmented Generation, AI agents, and multimodal applications using Transformer architectures.

At BMW Group (Research & Technology), I work on autonomous decision-making systems and scalable ML pipelines. Previously at Fraunhofer IGCV, I developed ML models for manufacturing and robotics. I also worked as a Software Engineer at Refresh-K LLC on database and document processing automation. I've served as a board member of UnitedAI club.



Email  /  LinkedIn  /  Github  /  Academic Writing

profile photo


Current technologies

I'm proficient in a range of modern technologies that empower me to build highly functional solutions. These are some of my main technologies.

AI & Machine Learning

PyTorch

ML framework

TensorFlow

ML platform

LlamaIndex

LLM data framework

Cloud & Databases

PostgreSQL

Relational database

MongoDB

NoSQL database

Docker

Containerization

AWS

Cloud platform

Azure

Microsoft cloud

Backend & Tools

Node.js

JavaScript runtime

Supabase

Backend tool

Django

Python web framework

Git

Version control

Frontend & Design

TypeScript

JavaScript but better

React

JavaScript Library

NextJS

React framework

Tailwind

CSS framework

Figma

Design Tool

My Projects


Autonomous Lawn Mower
Autonomous Lawn Mower RTK & Navigation
Developed a precision navigation system using RTK GPS with autonomous obstacle detection and avoidance capabilities.
C++ Raspberry Pi RTK GPS ROS Sensor Fusion Linux
[webpage]
Mines Detection with Drones
Mines Detection with Drones and Other Methods
Designed multi-sensor systems for humanitarian demining efforts with UnitedAI, combining aerial imagery and ground-based detection.
LiDAR Computer Vision Multispectral Imaging Radar Physics Simulation PyTorch
[webpage]
Mirobot 6DOF Mini Robotic Arm
Mirobot 6DOF Mini Robotic Arm
Implemented a vision-language model system enabling a robotic arm to execute tasks based on natural language instructions and visual input.
Vision-Language Models Reinforcement Learning Robotics RealSense PyTorch ROS
[webpage]
Turtlebot Maze Solver
Turtlebot Maze Solver (Sim & Real)
Created an autonomous navigation system for complex maze environments using LiDAR-based mapping and path planning, implemented in both simulation and real-world.
ROS LiDAR SLAM Path Planning Gazebo Python
[webpage]
NVIDIA Jetbot
NVIDIA Jetbot Gesture Recognition
Built a responsive robot that interprets hand gestures in real-time using vision-based deep learning for human-robot interaction.
PyTorch YOLO ROS NVIDIA Jetson Computer Vision TensorRT
[webpage]
CERN GridWars
CERN Spring Campus 2024 GridWars Hackathon
Competed in CERN's strategic AI programming hackathon, developing algorithms for resource management in simulated grid environments.
Java Algorithms Game Theory Strategic Planning Multi-agent Systems
[code] [certificate]
B-Plus Co-op
B-Plus Co-op Label Errors Detection
Developed an automated system to identify and correct annotation errors in autonomous driving datasets, improving model training accuracy.
Computer Vision PyTorch CARLA Simulator Data Cleaning Python OpenCV
[code] [webpage]
Easy Knowledge
Easy Knowledge
Created an AI-powered learning platform that personalizes educational content based on individual reading patterns and comprehension levels.
React Node.js GPT API LangChain TypeScript PostgreSQL GraphQL TailwindCSS
[code]
Neon Cube
Neon Cube
Designed and published a neon-themed puzzle-platformer game with procedurally generated levels, integrated analytics and monetization.
Godot Firebase GDScript Mobile Development Game Design Pixel Art
[itch.io] [webpage]
Board Games RAG
Board Games RAG
Built a specialized retrieval-augmented generation system for answering complex board game rule questions with contextually relevant information.
RAG Vector Databases Hugging Face Mistral Semantic Search Embeddings
[code] [webpage]
Vision Transformer
Vision Transformer from Scratch
Implemented a Vision Transformer architecture from first principles to understand self-attention mechanisms for image data.
Transformers PyTorch Deep Learning Computer Vision Self-Attention
Mensa Companion
Mensa Companion: University Canteen Speech Assistant
Developed a conversational AI system for navigating university dining options, with voice recognition and menu recommendation capabilities.
Rasa NLU Speech Recognition Conversational AI Python Docker
[code]
Semantic Search
Semantic Search with Transformers
Created a semantic search engine for scientific literature using transformer-based embeddings to find conceptually similar papers.
NLP Transformers Hugging Face Vector Search FAISS Python
[code]
Cooking Recipes
Cooking Recipes Management System
Built a scalable, containerized web application for recipe management with user authentication, favorites, and personalized recommendations.
Spring Boot Java MongoDB Docker Azure CI/CD RESTful API
[code]
Diophantine Equations
Diophantine Equations Solver
Developed a symbolic mathematics solver that parses and solves complex diophantine equations with formal grammar and constraint satisfaction.
Python Z3 Solver Formal Languages Constraint Programming Symbolic Math
[code]
Car Prices Prediction
Auction Car Prices Prediction
Created a desktop application that predicts auction car prices using machine learning, with interactive feature selection and result visualization.
Python Random Forest PyQt Pandas Scikit-learn Data Visualization
[code]
Snake Telegram Bot
Snake Telegram Bot
Built an image classification bot that identifies snake species from user-submitted photos with real-time feedback.
Python Telegram API OpenCV CNN Transfer Learning Image Classification
[code]
Gnosis ChatBot
Gnosis ChatBot
Developed a conversational ordering system with natural language understanding for pizza customization and order processing.
React Node.js NLU Azure Cognitive Services Intent Recognition Entity Extraction
[code]
Tic-Tac-Toe
Tic-Tac-Toe
Implemented a Tic-Tac-Toe game with multiple AI difficulty levels using minimax algorithm and alpha-beta pruning.
Python Game Theory Minimax Alpha-Beta Pruning GUI Development
[code]
SVD Image Compressor
SVD Image Compressor
Created a tool that compresses images using singular value decomposition, demonstrating the balance between compression ratio and quality.
Python Linear Algebra NumPy Image Processing Matplotlib Data Compression
[code]
Blackjack with RL
Solving Blackjack with Reinforcement Learning
Trained a reinforcement learning agent to optimize playing strategy in Blackjack through experience-based policy improvement.
Python Reinforcement Learning Q-learning OpenAI Gym Policy Optimization TensorFlow
[code]