Gulzhas Mailybayeva, AI/ML Engineer
Machine Learning · NLP · LLM Orchestration · RAG
I build and ship LLM-powered systems. I developed ai-test-gen -- a hybrid rule-based + LLM pipeline for test case generation that I actively maintain -- and a trading app (ai-gap-trading-forecaster) I use as an adversarial UI testbed.
2
AI Systems Shipped
1
Research Projects
8
Learning Topics
ML · Conversational · AWS
Courses
Learn
Machine Learning
Foundations of machine learning: supervised & unsupervised learning, model evaluation, feature engineering, and practical implementations with PyTorch and Scikit-learn.
Artificial Intelligence
Broad AI concepts: search algorithms, knowledge representation, planning, reasoning, and the landscape of modern AI systems.
Natural Language Processing
Text processing, language models, sentiment analysis, named entity recognition, and modern transformer-based NLP.
Statistical Planning and Reinforcement Learning
Planning under uncertainty, Markov decision processes, reinforcement learning algorithms, policy optimization, and multi-agent systems.
Tools & Frameworks
Flagship Systems
ai-test-gen
AI-Powered Test Case Generation Framework
A multi-agent, hybrid rule-based + LLM orchestration system that generates structured manual test cases from Azure DevOps user stories. Deterministic scaffolding handles structure, while Gemini-powered LLM correction and ChromaDB RAG enforce high-quality, consistent steps with automated upload to ADO Test Plans.
ai-gap-trading-forecaster
RL Forecasting Engine & Adversarial UI Testbed
A time-series trading application powered by RL-style decision logic and forecasting that doubles as an adversarial UI testbed. The frontend is intentionally volatile (mutating DOM, shifting locators) to create a realistic, hostile environment for validating the self-healing Playwright locators and LLM+RAG test generation pipeline in ai-test-gen.
Recent Posts
View allHow I Built a Hybrid Rule-Based + LLM Test Generation Pipeline
8 min870 test cases, $0.002 each, 92% time saved. A deep dive into combining deterministic scaffolding with Gemini 2.5 Flash for automated test case generation from user stories.
From 10 Years of Java to Python for AI: What Actually Transfers
6 minAfter a decade of Java enterprise development, here's what carried over to Python ML/AI work -- and what I had to completely relearn.