cv

Contact Information

Name Alafate Abulimiti
Email
Location Paris, France
Website https://neuromaancer.github.io

Professional Summary

AI researcher and engineer working on conversational AI, reinforcement learning and human–agent interaction. Postdoctoral research at Sorbonne University, following a PhD at ENS and applied ML for scientific R&D at L’Oréal; publications at ACL and SIGDIAL.

Work Experience

  • 2025 - Present
    Postdoctoral Researcher
    ISIR, Sorbonne University
    Human–agent collaboration, PEPR eNSEMBLE, with Catherine Pelachaud and Agnès Helme-Guizon.
    • Designed and implemented an AI agent that guides simulated group discussions about behavior change, using a discrete SAC policy to select dialogue acts.
    • Built LLM simulations of two participants; designed experiments comparing the policy with four LLM dialogue managers and analyzed adaptation across participant profiles (SIGDIAL 2026).
    • Develop agent training and evaluation environments with PyTorch, Hugging Face Transformers and vLLM. Current work includes PPO experiments and multiparty turn-taking research.
  • 2025
    Visiting Researcher
    University of Gothenburg
    Hosted by the CLASP team in October 2025.
  • 2024 - 2024
    AI Engineer / Data Scientist
    L’Oréal Advanced Research
    Interviewed scientists to identify R&D needs and help design three MVPs testing AI adoption: molecular prediction, document extraction and image enhancement.
    • Developed molecular-property prediction models for scientific R&D; this work was subsequently used in an internal tool.
    • Compared descriptor regressors, SMILES language models and graph models for melting points and Hansen solubility parameters (XGBoost, ChemBERTa, chemprop).
    • Prepared BigQuery/GCP molecular data with RDKit; evaluated with cross-validation and R²/MAE/RMSE, removing public-training/internal-evaluation molecule overlap.
    • Designed a GenAI information-extraction protocol for internal reports and external publications; also implemented LLM/RAG prototypes with LangChain and Vertex AI.
  • 2020 - 2023
    Doctoral Researcher
    École normale supérieure (ENS) / INRIA Paris, ALMAnaCH team
    Socially adaptive dialogue in peer tutoring. Supervisors: Justine Cassell and Chloé Clavel.
    • Built explainable NLP classifiers for off-task talk, combining linguistic and social features; analyzed rapport-related signals with SHAP.
    • Proposed a preliminary social-planning model representing task goals, social goals and rapport dynamics to guide off-task exchanges.
    • Developed hedge generation through language-model fine-tuning and classifier reranking; assessed outputs with automatic metrics, human evaluation and error analysis (ACL 2023, Long Papers).
    • Compared gradient-boosting and MLP/LSTM models for hedge timing using dialogue history, embeddings and nonverbal cues; checked feature contributions with ablations (SIGDIAL 2023).
  • 2020 - 2020
    Research Intern
    LLF, Université de Paris
    • Developed a formal dialogue model using Type Theory with Records (TTR), combining character and conversational types to study strategic choice; published at SemDial 2020.
  • 2019 - 2019
    Data Scientist Intern
    Bouygues
    • Developed a limited-data NLP prototype to extract and score skills from CVs to support internal CV screening, using a skills vocabulary, fastText/Gensim and one-class SVM.

Education

  • 2020 - 2023

    Paris, France

    PhD
    École normale supérieure (ENS)
    Artificial Intelligence research
    • Research focus: NLP, Transformers and human–computer interaction.
    • Thesis: socio-conversational strategies in peer tutoring, focusing on off-task talk and hedges.
  • 2019 - 2020

    Tours, France

    Master’s degree
    University of Tours
    Big Data Management and Analytics (BDMA)
    • Studies: NLP, deep learning, big data and recommender systems.
    • Project: exploring MWE identification with BERT, morphosyntax and BiLSTM.
  • 2016 - 2019

    Tours, France

    Engineering degree
    Polytech Tours
    Computer Science
    • Software engineering and object-oriented programming (Java/C++).
    • Graduation project: MLP/LSTM sequence prediction for operations research.
  • 2012 - 2016
    B.Sc.
    Beijing Institute of Technology
    Applied Physics
    • Mathematical modelling, linear algebra and statistics.

Selected Training

  • Reinforcement Learning Specialization - Coursera (2020)
  • Natural Language Processing Specialization - Coursera (2020)
  • Build a Neo4j-backed Chatbot using Python - Neo4j (2024)

Skills

Domains & Methods
NLP / LLM Human–agent interaction Reinforcement learning Multi-agent simulation Model evaluation Multimodal analysis Fine-tuning SHAP / interpretability
Programming & Libraries
Python PyTorch Hugging Face Transformers scikit-learn NumPy / Pandas SQL / PySpark
Tools & Platforms
vLLM LangChain BERT spaCy fastText/word2vec Gensim OpenFace/OpenSMILE TensorFlow XGBoost MLP/LSTM Linux / Bash GCP (basic) OVH Jean Zay Vertex AI BigQuery Git / GitLab / GitHub Docker CI/CD (GitLab CI) DVC MLflow W&B Matplotlib Seaborn UMAP
Soft Skills
Analytical thinking and problem solving Planning and coordination Interdisciplinary and multicultural collaboration Scientific communication, teaching and mentoring Scientific needs analysis and problem framing

Languages

French Fluent
English Fluent
Uyghur Native
Chinese Native

EF SET English: 74/100 (C2), 2020.

Interests

Chess chess passionate, started since 7yo
Meditation Vipassana, stress management, concentration
Reading philosophy, linguistics, social science, psychology
CrossFit just want to keep fit, stress relief

Scientific Activities