Available for collaboration

Hichem Telli

|

Researcher at Huawei Cloud R&D · Helsinki, Finland
Pioneering Generative AI for Autonomous Driving

4 h-index
65+ Citations
5+ Years Research
View Research Google Scholar
Scroll to explore
About Me

Building the Future of
Autonomous Perception

I am a Computer Vision and Generative AI researcher at Huawei Cloud R&D in Helsinki, Finland, where I push the boundaries of what machines can see, understand, and generate.

My research journey began with a PhD in Biometrics at the University of Biskra (Algeria), where I specialized in kinship verification and facial analysis using deep learning. Since joining Huawei in 2023, I've pivoted to large-scale industrial AI, working across OCR systems, Visual Language Models, Large Language Models, and most recently, Generative AI for Autonomous Driving.

Today, my primary focus spans Video Generation for Autonomous Driving — pushing the boundaries of 4D LiDAR video synthesis to create photorealistic, controllable, and temporally consistent sensor simulations for next-generation autonomous vehicles. I am currently working on a novel approach to multi-sensor Dual-Modality (LiDAR + Camera) video generation that is under active development and review.

Beyond autonomous driving, I am deeply passionate about Artificial General Intelligence — exploring abstract reasoning, program synthesis, and the path toward human-level AI. I actively engage with ARC-AGI v2 & v3 challenges, particularly the interactive variants, as a rigorous testbed for generalization and compositional reasoning.

Helsinki, Finland
Huawei Cloud R&D
GenAI · Autonomous Driving · LiDAR
PhD – University of Biskra, Algeria
Hichem Telli

Hichem Telli

CV Researcher @ Huawei

Helsinki, FI

Current Focus

Video Generation for Autonomous Driving

LiDAR Video Generation Diffusion Models Transformers ControlNet 3D-aware Generation ARC-AGI v3 (Interactive)

Current Focus

Artificial General Intelligence

ARC-AGI v3 (Interactive) Abstract Reasoning Program Synthesis Reinforcement Learning Neuro-Symbolic AI Reasoning Generalization
Research

Areas of Expertise

From biometric recognition to large-scale generative AI systems

Current Focus

Generative AI for Autonomous Driving

Developing controllable 4D Dual-Modality (LiDAR + Camera) video generation with diffusion models and transformer architectures. Enabling multi-sensor joint simulation for autonomous vehicle testing and development.

Diffusion Models Dual-Modality 4D Generation ControlNet LiDAR Simulation

Multimodal Vision-Language Models

Research on VLLMs and LLMs for document understanding, complex visual reasoning, and multimodal representation learning at scale.

VLLM Document AI Multimodal

Image & Video Generation

Generative models for high-fidelity image synthesis and controllable video generation, with applications in synthetic data creation for autonomous systems.

Image Generation Video Synthesis Controllability

Biometrics & Face Analysis

Deep learning for kinship verification, personality trait estimation from facial imagery, and robust face recognition using local and global descriptors.

Kinship Verification Face Recognition Biometrics

OCR & Document Intelligence (Multilingual)

Optical character recognition systems and intelligent document processing pipelines for industrial-scale document understanding workflows.

OCR Document Understanding IE

Large Language Models

Applied research on LLMs for reasoning, natural language understanding, and integration within multimodal AI pipelines.

LLMs Reasoning NLU
Publications

Selected Research Works

View full list on Google Scholar

CVPRW 2026 · Workshop 2026 🏆 2nd Place

Conflict-Aware Multimodal Fusion for Ambivalence and Hesitancy Recognition

Salah Eddine Bekhouche, Hichem Telli, Azeddine Benlamoudi, Salah Eddine Herrouz, Abdelmalik Taleb-Ahmed, Abdenour Hadid

A conflict-aware multimodal fusion approach for recognizing ambivalence and hesitancy, achieving 2nd place at the CVPR 2026 Workshop competition.

Multimodal Fusion Affective Computing Computer Vision Conflict-Aware
arXiv
CoRR / QIAS 2025 2025

CVPD at QIAS 2025 Shared Task: An Efficient Encoder-Based Approach for Islamic Inheritance Reasoning

Salah Eddine Bekhouche, Abdellah Zakaria Sellam, Hichem Telli, Cosimo Distante, Abdenour Hadid

An efficient encoder-based approach for the QIAS 2025 Shared Task on Islamic Inheritance Reasoning, demonstrating strong performance on legal reasoning benchmarks.

NLP Legal Reasoning Encoder Models
arXiv DBLP
Traitement du Signal · Journal 2021

A Novel Multi-Level Pyramid Co-Variance Operators for Estimation of Personality Traits and Job Screening Scores

Hichem Telli, Salim Sbaa, Salah Eddine Bekhouche, Fadi Dornaika, Abdelmalik Taleb-Ahmed, Miguel Bordallo López

A novel multi-level pyramid co-variance operator framework for automatic estimation of personality traits and job screening scores from visual data, achieving state-of-the-art results.

Personality Traits Affective Computing Co-variance
DOI DBLP
IC_ASET 2019 · IEEE Conference 2019

Deep Features for Kinship Verification from Facial Images

Hichem Telli, et al.

Leveraging deep convolutional neural network features for kinship verification from facial images, demonstrating robust family relationship detection across varied conditions.

Kinship Verification Face Recognition Deep Learning
IEEE Xplore
SIGPROMD 2018 · Workshop 2018

LPQ and LDP Descriptors with ML Representation For Kinship Verification

Hichem Telli, et al.

Combining Local Phase Quantization (LPQ) and Local Directional Pattern (LDP) descriptors with machine learning for robust kinship verification.

Local Descriptors LPQ LDP Kinship
ResearchGate
Experience

Professional Journey

Huawei Cloud R&D. Helsinki, Finland
June 2023 – Present

Computer Vision Researcher

Conducting industrial research across multiple advanced AI domains, progressing from perception to full generative AI systems.

2025 – Present
GenAI for Autonomous Driving

Researching controllable Dual-Modality (LiDAR + Camera) 4D video generation with diffusion transformers and multi-sensor architectures for autonomous vehicle simulation.

2025
Image & Video Generation

Generative models for synthetic image/video data creation and controllable generation for simulation pipelines.

2024
VLLMs & Large Language Models

Multimodal language models for document understanding and complex visual reasoning at industrial scale.

2023 – 2024
OCR Systems

Optical character recognition pipelines for large-scale document processing applications.

University of Biskra Biskra, Algeria
Sept. 2017 – 2023

PhD Researcher – Biometrics & Computer Vision

Doctoral research focused on biometric recognition systems, kinship verification, face analysis, and personality computing using deep learning. Published peer-reviewed work in IEEE conferences and international journals.

PyTorchDeep LearningMachine LearningBiometricsFace AnalysisCNN/RNN
University of Biskra Biskra, Algeria
2011 – 2013

M.Sc. in Advanced Automatics

Master's degree in Advanced Automatics, building a strong foundation in control systems, signal processing, and applied mathematics.

RoboticsControl SystemsSignal ProcessingApplied MathematicsMachine LearningFuzzy Logic
Skills

Technical Toolkit

Deep Learning & AI

Diffusion Models Transformers ControlNet CNN / RNN / LSTM VLLMs / LLMs Reinforcement Learning Self-Supervised Learning Transfer Learning

Autonomous Driving

LiDAR Simulation 4D Video Generation Controllable Generation Temporal Consistency nuScenes Waymo Dataset Proprietary Datasets Synthetic Data

Programming & Frameworks

Python PyTorch C / C++ Java PHP / Laravel MATLAB NodeJS

Computer Vision

Image Generation Video Generation Object Detection Face Recognition OCR Image Segmentation Object Tracking Pattern Recognition

Tools & Infrastructure

CUDA / GPU / Ascend NPU Git Docker Linux / Ubuntu LaTeX Markdown REST APIs / GraphQL Kaggle Slurm ModelArts

Languages

Arabic (Native) French (Intermediate) English (Fluent)
Contact

Let's Connect

I'm always open to research collaborations, discussions, and opportunities.