Михаил

Mykhailo Kohutenko
AI & Infrastructure Engineer
✉ [відкрити контакти](див. вище в блоці «контактна інформація») ☎ [відкрити контакти](див. вище в блоці «контактна інформація») linkedin.com Kyiv, Ukraine

SUMMARY
AI & Infrastructure Engineer with a strong mathematical foundation (BSc in Mathematics & Computer Science)
and an ongoing MSc in Artificial Intelligence Systems. Experienced in designing production-grade GraphRAG
architectures, asynchronous Multi-Agent LLM systems, and deploying physical AI compute clusters for
enterprise environments. Reduced LLM token consumption by 93x through topological AST-based code
analysis and automated L1/L2 engineering workflows via custom GenAI microservices.

TECHNICAL SKILLS

Programming Languages Python (Core, OOP, AST), SQL, Cypher
GraphRAG, LLM Orchestration, Gemini API, Groq API, OpenRouter, Vector
AI & LLM Engineering:
Embeddings, Prompt Engineering, Transformers, Hugging Face
Deep Learning (CNN, RNN, LSTM, GRU), PyTorch, TensorFlow, Scikit-learn,
Machine Learning
NumPy, Pandas
Backend Development FastAPI, asyncio, aiogram 3.x, REST APIs, JSON (Schema Serialization)

Databases Neo4j (Graph), SQLite, aiosqlite, Relational Schema Design

Virtualisation & Infra Docker, Linux, VMware, Bare-Metal Servers
AI Cluster Initialization & Configuration, Server Hardware Validation,
Networking & Hardware
TCP/IP, VLANs
Tools & Source Control Git, Jupyter Notebooks
Graph Theory, Linear Algebra, Mathematical Statistics, Probability Theory,
Mathematical Foundations
Optimization Methods, Genetic Algorithms

PROFESSIONAL EXPERIENCE
Alesta UA 10/2024 – Present
AI & Infrastructure Engineer
Kyiv, Ukraine
1. Enterprise GraphRAG Code & Configuration Analyzer (Internal R&D)
Project Description: An internal enterprise tool designed for the engineering department to analyze large-
scale infrastructure repositories and automation scripts. The system reduces the time required for code
review and architectural diagnostics by performing semantic and topological search across complex code
dependencies.
Responsibilities and Key Accomplishments:
• Developed the backend module for parsing Python source code into Abstract Syntax Trees (AST) and
extracting structural relationships.
• Designed and populated the Neo4j graph database schema to map code dependencies (calls, imports,
inheritance) using Cypher queries.
• Implemented a hybrid retrieval pipeline combining topological subgraph traversal with Gemini API
vector embeddings, which reduced the LLM context token usage by 93x.
• Collaborated with the infrastructure team to wrap the analytical engine into high-concurrency RESTful
endpoints using FastAPI.
Technology Stack: Python, Neo4j, Cypher, FastAPI, Gemini API, Vector Embeddings, AST, JSON, Docker.
2. Enterprise Multi-Agent AI Copilot for Engineering Support
Project Description: A corporate asynchronous Telegram-based GenAI assistant built for L1/L2 tech
support and field engineers. The product provides the technical team with instant, context-aware access
to hardware troubleshooting protocols, manuals, and server diagnostic guidelines directly from the data
center floor.

Responsibilities and Key Accomplishments:
• Built the core asynchronous backend architecture utilizing Python (asyncio, aiogram 3.x) and Finite
State Machine (FSM) routers to handle concurrent multi-user sessions.
• Integrated a multi-provider LLM routing layer (Groq API, OpenRouter) via JSON REST APIs to ensure
high availability and fallback routing.
• Designed the relational database schema using SQLite (aiosqlite) for session state persistence,
dynamic prompt injection, and audit logging.
• Connected a real-time multimodal voice synthesis module (edge-tts) to allow hands-free interaction
for engineers working with hardware.
Technology Stack: Python, asyncio, aiogram 3.x, SQLite, aiosqlite, Groq API, OpenRouter, edge-tts, REST
API, JSON.
3. AI Cluster & Server Infrastructure Deployment
Project Description: B2B infrastructure integration projects for enterprise and government clients. The
focus is on preparing, deploying, and validating high-availability bare-metal clusters and storage solutions
optimized for heavy machine learning and distributed computing workloads.
Responsibilities and Key Accomplishments:
• Participated in hardware selection, capacity planning, and bare-metal initialization for compute-
intensive ML environments.
• Configured base server environments, including operating system provisioning, basic network
segmentation (VLANs, TCP/IP), and hypervisor setup.
• Conducted system inspections, stress-testing, and validation of hardware components before final
client handover.
• Assisted senior architects in drafting technical documentation, architectural specifications, and
deployment reports for client CTOs.
Technology Stack: Bare-Metal Servers, Linux, Docker, CUDA toolkit, Virtualization (VMware), TCP/IP,
VLANs.

EDUCATION
MSc in Systems and Methods of Artificial Intelligence 09/2025 – 12/2026
Igor Sikorsky Kyiv Polytechnic Institute (KPI), IASA
Kyiv, Ukraine
• Master's Thesis: Topological Code Analysis and Context Optimization using GraphRAG, AST Parsing,
and Vector Embeddings.
• Applied Research: Implemented neural network optimization algorithms (Backpropagation, Genetic
Algorithms) and graph-based data structures in Python.

BSc in Mathematics and Computer Science 09/2021 – 05/2025
V. N. Karazin Kharkiv National University
Kharkiv, Ukraine
• Core Focus: Linear Algebra, Mathematical Statistics, Optimization Methods, and Probability Theory.

CERTIFICATES
Natural Language Processing Specialization Deep Learning Specialization
DeepLearning.AI & Stanford University DeepLearning.AI & Stanford University
(Built sentiment analysis, NER, and transformer- (Engineered deep L-layer neural networks from
based sequence models using RNNs, LSTMs, scratch in NumPy and CNNs in TensorFlow for
GRUs, and Hugging Face). image recognition).

Machine Learning Specialization
DeepLearning.AI & Stanford University
(Implemented supervised regression /
classification models with L1/L2 regularization
and collaborative filtering recommender systems).