Andrii Bashtovyi

Senior AI Developer

01

Education

Master's degree in Mathematics and StatisticsTaras Shevchenko National University of Kyiv
02

Machine Learning Engineer

R&D Institute Ukraine
Project scope under NDA

Built a generative neural network that outperformed existing open-source solutions in its field, leading to a paper submission to ACM CCS 2022.

  • Deep neural networks
  • Patent preparation
  • Graph databases
  • Data engineering
03

Senior AI Developer

Enterprise search, an internal GPT assistant, and production standards for AI across teams.

  • Azure GPT
  • OpenAI
  • RAG
  • PySpark
03.1

Intelligent Search — Enterprise Search with GPT

Ontario Power GenerationProject 1 of 3

Led the data engineering team building an end-to-end search and document pipeline: ingestion, preprocessing, summarisation and vectorisation across structured and unstructured enterprise data.

  • OpenAI embeddings
  • Azure Form Recognizer
  • PySpark
03.2

ChatOPG — Internal RAG & GPT Assistant

Ontario Power GenerationProject 2 of 3

Supervised development and fine-tuning of an internal Azure GPT and RAG assistant that follows organisational knowledge and content policies. Led the intent classification and routing that sends each request to a domain-specific assistant.

  • Azure GPT
  • RAG
  • Intent routing
03.3

Engineering leadership

Ontario Power GenerationProject 3 of 3

Managed AI model deployment and production practices, set development and coding standards, consulted across teams, and took part in technical interviews.

04

Senior AI Developer

GenTech

Production LLM systems: a GraphRAG customer support chatbot and an enterprise knowledge base.

  • GPT-4 Turbo
  • GraphRAG
  • PostgreSQL
  • RabbitMQ
04.1

LLM & GraphRAG Customer Support Chatbot

GenTechProject 1 of 2

Led the second development phase and production deployment of a large-scale GPT-4 Turbo support chatbot. Added GraphRAG for retrieval across connected sources, plus tool calling, relational databases, custom APIs and RabbitMQ for high-throughput processing.

  • Tool calling
  • Custom APIs
04.2

Knowledgebase — Enterprise RAG & GraphRAG

GenTechProject 2 of 2
+20%semantic search relevance

Built an in-house PostgreSQL RAG platform over corporate knowledge sources, including support databases and Google Drive. Implemented Microsoft GraphRAG over unstructured documents, improving retrieval from complex PDF and Word files.

  • Microsoft GraphRAG
  • Google Drive
05

Project-scoped work

WebLegends
05.1

Extractor & Template Generator

WebLegendsProject 1 of 2

Production document extraction on multimodal LLMs for large PDFs with complex, nested tables. Benchmarked Gemini 2.5, Claude 4, GPT-o3 and Llama 4 Maverick for accuracy, cost and latency, and added prompt caching. A companion Template Generator turns arbitrary forms into structured extraction schemas.

  • Multimodal LLMs
  • Prompt caching
05.2

AI Salesman

WebLegendsProject 2 of 2

Multi-agent sales assistant for a car-dealership platform on the OpenAI Assistants API, with hybrid-search RAG over current vehicle and deal data.

  • OpenAI Assistants
  • Hybrid search
06

Senior AI Developer

Industrial AI for energy operations: a multi-tenant assistant, a real-time agentic decision system, and automated risk reporting.

  • GPT-6
  • Gemini 3.5 Flash
  • LangSmith
06.1

Industrial AI Assistant — Multi-tenant Text-to-SQL & RAG

CleanConnect.aiProject 1 of 3
~73%smaller prompts

Core components of a multi-tenant assistant answering natural-language questions over real-time industrial, facility and financial data. A unified classification layer routes each request to multi-agent text-to-SQL, RAG with web search, or general handling.

  • Parallel LLM execution
  • Automated evals
06.2

Real-Time Monitoring & Agentic Decision System

CleanConnect.aiProject 2 of 3

Event-driven agent that analyses live sensor data, combining detection events with ML forecasts to keep operational state current, assess emerging risks and trigger actions. Guardrails and decision logging keep it controllable and auditable.

  • Event-driven agents
  • ML forecasting
06.3

Automated Risk & Intelligence Reporting

CleanConnect.aiProject 3 of 3

Pipeline that turns multi-year time-series detection data into customer-facing operational risk reports with AI insights: composite risk indices, trend analysis and interactive HTML dashboards.

  • Time series
  • Risk indices