Portrait of Elgun Hasanov

01 / Profile

Elgün,if your keyboard allows it :)

I study how AI reasons, generalizes and earns trust

02 / Experience

How I got here

Education

2025 — 2027

Data and Artificial Intelligence

Master of Science

2022 — 2025

Computer Science

Bachelor of Science · Valedictorian

Research experience

Jan — Apr 2026

Student Researcher

NLP & Graph Methods

DIG Team · LTCI Lab
May — Sep 2025

Research Intern

Computer Vision & Deep Learning

AND Team · DISC Dept
Jan — May 2025

Research Intern

Unsupervised Learning

IMAGeS Team · D-IRTS Dept

Industry experience

Mar 2023 — Present

Co-Founder & Applied AI

RAG & Voice AI

Founding Team · Codage Tech
Apr — Sep 2026

Applied AI Engineer

LLM Fine-Tuning & Evaluation

Model R&D · AITOM
Jun 2024 — May 2025

Data & AI Architect Intern

Multi-agent AI & Data Platforms

Technology CoE · PASHA Life

Selected honors & awards

Nov 2025 · Finalist

Stanford ASES France Fellowship

HEC Paris deep-tech finalist among 300+ teams

Sep 2025 · BGF Scholarship recipient

French Government Excellence Scholarship

Awarded for my MSc studies at Polytechnique Paris

Jun 2024 · 1st place

Google AI Hackathon

FraudShield AI detects financial fraud with RAG

Nov 2024 · 1st place

GreenTech ClimAccelerator / COP29

Ecoffin pairs engine fuel with a CO₂ dashboard

More awards
May 2025 · Double winner

PASHA Hackathon 2.0 & 3.0

Invited as Innovation Mentor for 4.0, 5.0 & 6.0

Sep 2024 · Finalist

TEKNOFEST T3 AI Hackathon

Fine-tuned LLMs for Turkish-language understanding

11+ hackathon wins, that was fun :))

03 / Research

Publications & Preprints

Primary-Preserving Complementary Module Retrieval Across Formal Mathematics Libraries

  • MathComp
  • mathlib
Details & authors

Elgun Hasanov, Fatulla Bashirov, Nahla El Kadhi

Accepted for proceedings · Formal mathematics / semantic retrieval

Retrieval between MathComp/Rocq and mathlib/Lean is not always a one-to-one matching problem. This work studies complementary module retrieval while preserving the primary result, accounting for content distributed across different library boundaries.

Methods
Semantic retrieval · Graph methods · MathComp / Rocq · mathlib / Lean

Source reference

International Conference on Web Information Systems Engineering (WISE 2026) · Student Track

PaperCode

Asymmetric Cross-Cohort Generalization in Alzheimer's Disease Classification: A Bidirectional External Validation Study

  • External validation
  • ADNI / OASIS
Details & authors

Elgun Hasanov, Ayman El Kadhi, Gulzar Safarli, Nahla El Kadhi

Accepted full paper · Biomedical AI / external validation

A bidirectional external-validation study of Alzheimer's disease classifiers across independent clinical cohorts.

Methods
External validation · Distribution shift · Biomedical AI

Source reference

IEEE International Symposium on Computer-Based Medical Systems (CBMS 2026)

PaperCode

Leakage-Safe Machine Learning for Alzheimer's Disease Classification with External Validation on ADNI and OASIS

  • Leakage-safe
  • Clinical ML
Details & authors

Elgun Hasanov, Ayman El Kadhi, Gulzar Safarli, Nahla El Kadhi

Accepted for oral presentation and proceedings · Paper 337

A leakage-safe machine-learning evaluation across ADNI and OASIS with external validation.

Methods
External validation · Distribution shift · Biomedical AI

Source reference

KES International Conference

PaperCode

AttentionDep: Knowledge-Infused Attention for Interpretable Depression Severity Assessment from Social Media

  • Knowledge graph
  • Attention
Details & authors

Yusif Ibrahimov, Tarique Anwar, Tommy Yuan, Turan Mutallimov, Elgun Hasanov

Accepted for presentation · Mental-health NLP / interpretable attention

A knowledge-infused attention approach for interpretable depression-severity assessment from social-media text.

Methods
Attention · Knowledge infusion · Mental-health NLP

Earlier version available as an arXiv preprint.

Source reference

International Conference on Web Information Systems Engineering (WISE 2026) · Main Research Track

PaperCode

Can We Trust LLMs for Mental Health-Based Decisions? A Causality Aware Reliability Analysis

  • LLM reliability
  • Causal analysis
Details & authors

Yusif Ibrahimov, Turan Mutallimov, Seymour Mirzabayov, Elgun Hasanov

Published · Open access · LLM reliability / causal analysis

An analysis of Qwen2.5-7B's attention in mental-health classification examines whether predictive signals align with domain reasoning. The work identifies a limitation in attention-based causal discovery and studies a correction using contrastive TF-IDF and directed pointwise mutual information.

Methods
Causal analysis · Trustworthy AI · Qwen2.5-7B

Artificial Intelligence for Digital Transformations (AIDT) · pp. 251–263 · DOI: 10.1007/978-3-032-31319-5_17

PaperCode

Decision Support through Feature-Aware Ensemble Clustering: A Multi-Algorithm Fusion Framework for Complex Data Analysis

  • Ensemble clustering
  • Decision support
Details & authors

Elgun Hasanov, Nahla El Zant El Kadhi

Published · Ensemble clustering / decision support

The framework combines complementary clustering signals while accounting for feature structure, aiming to produce more robust groupings for complex data-analysis workflows.

Methods
Ensemble clustering · Decision support

International Conference on Decision Aid Sciences and Applications (DASA) · pp. 980–986 · DOI: 10.1109/DASA68193.2025.11498998

PaperCode

Rapid Automated Alzheimer's Disease Classification Using FastSurfer Segmentation and Explainable Machine Learning

  • FastSurfer
  • Explainable ML
Details & authors

Gulzar Safarli, Ayman El Kadhi, Elgun Hasanov, Nahla El Zant El Kadhi

Published · Medical imaging / explainable ML

This project connects FastSurfer-derived neuroanatomical features with interpretable machine-learning models to support rapid Alzheimer's disease classification.

Methods
FastSurfer · Explainable ML · Brain segmentation

International Conference on Decision Aid Sciences and Applications (DASA) · pp. 1527–1535 · DOI: 10.1109/DASA68193.2025.11499040

PaperCode

AttentionDep: Domain-Aware Attention for Explainable Depression Severity Assessment

  • Domain attention
  • Mental-health NLP
Details & authors

Yusif Ibrahimov, Tarique Anwar, Tommy Yuan, Turan Mutallimov, Elgun Hasanov

Preprint · Earlier version of the WISE 2026 AttentionDep paper

A domain-aware attention model that combines contextual text representations with mental-health knowledge for interpretable, ordinal depression-severity assessment.

Methods
Attention · Knowledge infusion · Mental-health NLP

Earlier version of the WISE 2026 AttentionDep work.

arXiv preprint arXiv:2510.00706 · DOI: 10.48550/arXiv.2510.00706

PaperCode

*conceptual previews, powered by imagination :)

LEGO style portrait of Elgun HasanovBack to top
Contact