👋 Hi, I’m Naz
AI Engineer | Data Scientist | Python Developer
I build AI systems that transform research ideas into scalable software solutions.
With a background in Artificial Intelligence and Computer Science, I specialize in designing, building, and deploying machine learning and LLM-powered applications.
What I Do
- Build end-to-end ML & AI systems
- Develop LLM & Retrieval-Augmented Generation (RAG) applications
- Build evaluation and observability pipelines
- Design scalable APIs for AI inference
- Apply data science & machine learning to real-world problems
Technical Skills
Languages: Python (Advanced), SQL, Java, C/C++, R
AI/ML: LangChain, LangGraph, LlamaIndex, PyTorch
Agentic AI: CrewAI, BeeAI, AutoGen
LLM Platforms and APIs: OpenAI API, Ollama, Hugging Face, AWS Bedrock
Vector Databases and Retrieval: ChromaDB, FAISS, Hugging Face Embeddings
Backend and Deployment: FastAPI, FLask, REST APIs , GitHub Actions, CI/CD fundamentals
Frontend prototyping: Streamlit, Gradio
Evaluation and Observability: RAGAS, LangSmith, LangFuse, Logging, MLFlow
Tools: Linux, Jupyter, Cursor, VS Code
Featured Projects
Production-Ready RAG Assistant
Tech: Python, FastAPI, Chroma DB, Hugging Face Transformers, RAGAS, MlFlow
- Built a scalable, modular Retrieval-Augmented Generation (RAG) system with FastAPI for intelligent document Q&A, featuring advanced chunking, hybrid retrieval (semantic + BM25), cross-encoder reranking, and comprehensive evaluation using RAGAS metrics.
- Included MLflow tracking, ChromaDB storage, a Streamlit dashboard, and support for multiple document formats, processing 10,000+ chunks efficiently with 4 core evaluation metrics for answer quality assessment.
visit rag repository
Small Language Model Evaluation Dashboard
Tech: Python, FastAPI, Ollama
- Implemented a small language model (SLM) evaluation dashboard that runs entirely offline using Ollama as the local inference backend. * Benchmarked multiple small models (2–5B parameters) across three phases: raw inference performance, structured output validation, and temperature variance analysis. Results are written to CSV and exposed via a REST API for a React dashboard.
- The following SLMs were benchmarked (the list is expanding): phi3-mini, gemma2-2b, qwen2.5-3b, llama3.2-3b
visit slm repository
End-to-End Machine Learning Pipeline - Time - Series Forecasting and Model Evaluation
Tech: Python, Scikit-learn, Pandas, XGBoost, LSTM, SARIMAX, Prophet
- Built a complete ML workflow from data ingestion to evaluation
- Compared multiple models using cross-validation
- Implemented experiment tracking & reproducibility
visit time-series repository
YouBot: AI-powered YouTube Video Summarizer and Q&A
Tech: LangChain, FAISS, Streamlit, HuggingFace Transformers
- Designed a RAG system using LangChain and HF embeddings
- Implemented request validation & logging
- Developed an AI chatbot for document querying and a separate pipeline for video transcript analysis
visit youbot repository
Currently working on
Conversational Agent with Memory + Guardrails
Tech: LangChain, LangGraph, LangSmith, Redis, ChromaDB, FastAPI, Pydantic, LangChain Embeddings, OpenAI API
- Implementing a multi-turn customer support agent with short-term (buffer) and long-term (vector database) memory, topic guardrails and fallback handling. Logs every failure to LangSmith.
Real-Time Streaming AI Assistant with Source Attribution Framework
- Working on a real-time AI response system using FastAPI and SSE to stream LLM-generated answers token-by-token to users. Building a citation and traceability framework that injects inline source references (document title, page, chunk ID) into generated responses.
Education
MSc Artificial Intelligence - University of Edinburgh, UK (2017)
BSc Computer Science - University College Dublin, Ireland (2013)
Let’s Connect
⭐ I’m actively seeking opportunities to contribute to AI-driven products and engineering teams.
Certifications
Previous Projects
- Time-Series Forecast
Time-Series EDA + Forecasting - Predicting Caribbean Climate for 2026 - 2027. The four different popular ML models (SARIMAX, XGBoost, LSTM and Prophet) were trained on Caribbean temperature anomalies and precipitation spanning 1980 - 2025, with Prophet outperforming the rest with MAE = 0.17C. The data was obtained from the NOAA historical dataset.
- Recommender System + Streamlit
The AI-powered movie recommender uses SVD collaborative filtering on the MovieLens 100K dataset to predict personalized movie ratings. It delivers top-N recommendations with confidence scores and generates creative taglines using Hugging Face FLAN-T5 offline. An interactive Plotly network graph visualizes user taste. Built with Streamlit.
Work Experience
Instructor @ Computer Science Department, Nazarbayev University [2018 - Present]
- Designed and taught Data Structures and Algorithms, Programming for Scientists and Engineers, Programming Languages, Research Methods using C, C++, Python, Java to over 1000 undergraduate students.
- Created syllabi, coding assignments and exams aligned with industry trends.
- Provided one-on-one guidance during office hours to troubleshoot technical and coding issues.
- Served on curriculum committees for Research Methods, hiring committees, and organised LT&T coding workshops to enhance student participation at competitions.
Research Assistant (Robotics and Machine Learning) @ Robotics and Mechatronics Department, Nazarbayev University [2014 - 2018]
- Assisted in design, implementation and testing of robotic systems, focusing on areas such as robot manipulation, control algorithms and human-robot interaction.
- Gathered and processed experimental data such as sonsor reading, simulation outputs utilizing tools like MATLAB and ROS.
- Performed comprehensive reviews of current Machine Learning, Aritificial Intelligence and Robotics literature and contributed to research publications, research and grant proposals.
- Maintained robotics lab equipment and hands-on guidance to undergraduate students in courses like Microcontrollers, Electrical and Electronic Circuits, Robotics and Research Methods.
Project Engineer Intern @ Microsoft [2013]
- Developed enterprise components and automated workflows using C# and SharePoint.
- Queried and validated structured datasets using SQL.
Research Projects @ Nazarbayev University
- SLIRS: Sign Language Interpreting Robotic System [2016]
- Adaptive Control Architecture for Humanoid Robotics based on Recurrent Neural Networks [2015]
- Development of an Intelligent Assistive Robot Manipulation System for Improving the Quality of Life of Disabled People [2015]
Relevant Publications
- Tazhigaliyeva, Nazgul, et al. “Cyrillic manual alphabet recognition in RGB and RGB-D data for sign
language interpreting robotic system (SLIRS).” Robotics and Automation (ICRA), 2017 IEEE
International Conference on. IEEE, 2017.
- Tazhigaliyeva, Nazgul, et al. “SLIRS: Sign language interpreting system for human-robot interaction.”
2016 AAAI Fall Symposium Series. 2016.
- Folgheraiter, Michele, Nazgul Tazhigaliyeva, and Aibek Niyetkaliyev. “Adaptive joint trajectory
generator based on a chaotic recurrent neural network.” Development and Learning and Epigenetic
Robotics (ICDL-EpiRob), 2015 Joint IEEE International Conference on. IEEE, 2015.
- Secured more than $100,000 in scholarships for academic achievements.
- Authored influential research publications in AI and HRI, accumulating 50+ citations.
- Volunteered with a local school’s Robotics Club, facilitating student participation in international robotics competitions.