Prompt Engineering for NLP with AI-IDE

2026/2027

Учебный год

ENG

Обучение ведется на английском языке

3

Кредиты

Кто читает:

Департамент политологии и международных отношений

Статус:

Курс по выбору

Когда читается:

2-й курс, 1 модуль

Преподаватель

Кольцов Сергей Николаевич

Course Syllabus

Abstract

The course “Prompt Engineering for NLP with AI-IDE” focuses on modern methods for developing and applying intelligent natural language processing systems using large language models (LLMs) and AI-oriented development environments. It covers the complete lifecycle of working with textual data, from basic preprocessing and classical machine learning methods to transformer architectures, local language models, and agent-based approaches to data analysis. The course examines prompt engineering as a key tool for controlling LLM behavior, including instruction-based, zero-shot, few-shot, and Chain-of-Thought prompts, as well as methods for improving the robustness and quality of generated outputs. Particular emphasis is placed on practical work in AI-IDEs (Cursor, Qoder, Visual Studio Code, Jupyter Notebook, and Codex) and on integrating language models into NLP pipelines.

Learning Objectives

• The course is designed to develop practical skills in building NLP projects with AI tools. It is intended for students and professionals interested in applying modern language models to the analysis and generation of textual data. The course comprises six thematic modules, each consisting of a lecture and a practical seminar. Throughout all modules, prompt engineering and AI-IDEs serve as cross-cutting tools for task specification, code generation and verification, data analysis, and result evaluation.

Expected Learning Outcomes

• explores the NLP tool ecosystem (scikit-learn, PyTorch, and Hugging Face), methods for building embeddings, the application of modern LLMs, and the deployment of local models using the Ollama framework
• knows agent-based systems built on locally deployed Qwen-family models for data analysis and the automation of research tasks

Course Contents

• Lecture 1. Introduction to Prompt Engineering and AI-IDE
• Lecture 2. NLP Tasks and AI-IDE-Assisted Text Data Preparation
• Lecture 3. Prompt-Driven ML Pipelines and Embeddings for NLP
• Lecture 4. Transformers and Prompt-Engineering Strategies for LLMs
• Lecture 5. Local LLMs and Prompt Engineering with Ollama
• Lecture 6. Agentic NLP Systems and Research-Task Automation

Assessment Elements

• Homework
• Exam
  The exam is conducted as a hackathon. All students receive a labeled dataset and must design and implement a machine learning algorithm that achieves the best possible performance across all classes. In particular, the F1-scores should be as balanced as possible across the classes. Students must present the results of their work. The presentation must include the following sections: 1. Project objective. 2. Description of the source data. 3. Algorithms used and their justification. 4. Model evaluation methodology. 5. Final results. Each presentation is allocated 10 minutes: 7 minutes for the presentation and 3 minutes for questions. Every student completes an individual project. The presentation and source code in the form of a Jupyter Notebook must be emailed to the instructor before the exam: skoltsov@hse.ru.

Interim Assessment

• 2026/2027 1st module
  0.5 * Homework + 0.5 * Exam

Bibliography

Recommended Core Bibliography

• Антонио, Д. Библиотека Keras – инструмент глубокого обучения. Реализация нейронных сетей с помощью библиотек Theano и TensorFlow / Д. Антонио, П. Суджит , перевод с английского А. А. Слинкин. — Москва : ДМК Пресс, 2018. — 294 с. — ISBN 978-5-97060-573-8. — Текст : электронный // Лань : электронно-библиотечная система. — URL: https://e.lanbook.com/book/111438 (дата обращения: 00.00.0000). — Режим доступа: для авториз. пользователей.
• Гудфеллоу, Я. Глубокое обучение / Я. Гудфеллоу, И. Бенджио, А. Курвилль , перевод с английского А. А. Слинкина. — 2-е изд. — Москва : ДМК Пресс, 2018. — 652 с. — ISBN 978-5-97060-618-6. — Текст : электронный // Лань : электронно-библиотечная система. — URL: https://e.lanbook.com/book/107901 (дата обращения: 00.00.0000). — Режим доступа: для авториз. пользователей.

Recommended Additional Bibliography

• Глубокое обучение на Python.. - 978-5-4461-0770-4 - Шолле Франсуа - 2021 - Санкт-Петербург: Питер - https://ibooks.ru/products/376833 - 376833 - iBOOKS

Authors

• Koltsov Sergei Nikolaevich