AI research · product · public expertise
Maria
Molchanova
Internationally experienced artificial intelligence researcher at MIRIAI and technology entrepreneur.
I research and build LLM-based systems, including RAG, Prompt & Instruction Engineering, conversational AI, personalization, data-centric AI, and quality evaluation. I publish findings at international venues, speak to research and industry audiences, and develop my own AI product.
Highlights
The essentials in 20 seconds
National television, a talk at the Russian Ministry of Science and Higher Education, recognition for an independent AI project, and a paper accepted at a workshop co-located with a major NLP conference.
Talk at Russia's Ministry of Science and Higher Education
AI and emerging technologies in education, science, and business.
Moscow Innovator 2026 nominee
My AI project passed the second stage of evaluation for the Moscow Mayor's award.
Paper accepted at GroundLM @ EMNLP 2026
RAG for small language models, grounding, and in-context distillation.
Research + Product
MIRIAI researcher; former Data Science Product Owner at Sber / Domclick; technology entrepreneur.
Research
From RAG to LLM Personalization
These topics reflect actual research projects, publications, and ongoing work rather than fashionable keywords.
RAG & Grounded AI
Retrieval-Augmented Generation, grounding, helping small language models use external knowledge, and improving the factual reliability and efficiency of AI systems.
Prompt & Instruction Engineering
Systematic design of LLM instructions: from prompt patterns and complex-task structuring to experimental evaluation.
Data-centric AI & Data Annotation
Gold datasets, crowdsourcing, expert annotation, LLM-assisted labeling, quality control, and inter-annotator agreement.
Conversational AI & AI Assistants
Dialogue-system architectures, AI assistants, speech functions, context management, and interaction scenarios.
Personalization & Computational Personality
Big Five, text-based personality prediction, simulation of personality traits, and controlled behavior of LLM personas.
LLM Evaluation, Synthetic Data & LLM-as-a-Judge
Generation quality evaluation, automated judge models, synthetic-data filtering, faithfulness, and groundedness.
Scientific Publications
International proceedings, ACL, IEEE, Springer, and EMNLP workshops
Published and publicly available research papers, with methods, results, and venue details.
Enhancing Small Language Models in RAG via Triplet-Guided In-Context Distillation
A training-free approach to RAG for small language models: an LLM creates structured examples that help an SLM use retrieved sources more effectively, without additional fine-tuning.
ChatGPT vs. Crowdsourcing vs. Experts: Annotating Open-Domain Conversations with Speech Functions
A comparison of experts, crowdworkers, and ChatGPT on complex discourse annotation. A structured decision-tree approach enabled an LLM to approach human-level performance on certain tasks.
Improving LLM-Driven Data Annotation Through Structured Instruction Engineering
A study of how instruction structure affects LLM annotation quality, treating instruction engineering as a controllable component of automatic annotation.
Exploring the Potential of Large Language Models for Text-Based Personality Prediction
Using LLMs for text-based personality prediction: the Big Five, prompt engineering, and comparisons with traditional approaches.
Text-based Personality Prediction Using Large Language Models
Further research on LLMs and personality: identifying personality traits from text and analyzing the effectiveness of prompt-based methods.
Exploring the Potential of Large Language Models to Simulate Personality
From recognizing personality to simulating it: can LLMs consistently reproduce target Big Five traits, and how can this be measured?
Talks
National and International Venues
The home page highlights selected events. The full list is ordered by prominence; international conference pages include photographs from the talks.
Talk at Russia's Ministry of Science and Higher Education
A presentation at the Ministry of Science and Higher Education of the Russian Federation on the development and practical applications of AI and emerging technologies.
The 5th IEEE International Conference on Human-Machine Systems
In-person research talk: “Improving LLM-Driven Data Annotation Through Structured Instruction Engineering.”
The 2nd International Conference on Foundation and Large Language Models
In-person talk: “Text-based Personality Prediction Using Large Language Models.”
Data Fest 2024 · Open Data Science (ODS)
A technical talk on data annotation, quality control, and annotation automation with LLMs.
Media
When Research Reaches Beyond the Lab
Expert commentary in national and business media, and coverage of scientific research.
Channel One · Good Morning
“Artificial Intelligence, You Made a Mistake!”
Kommersant Science
“Artificial Intelligence Deciphers the Human Mind”
Izvestia
“The Illusion of Independence: AI May Learn From Text Without Human Assistance”
Education