MS CS @ Rice University

.Hamza ElMokhtar Shili

I work on efficient and decentralized Mixture-of-Experts, model compression, and federated ML systems.

I am an M.S. student in Computer Science at Rice University , currently working on my thesis in scalable and federated ML systems fo text generation advised by Prof. Anastasios Kyrillidis. I like problems that sit at the boundary between algorithms and infrastructure: how to route, compress, and deploy models so they stay efficient and reliable in the wild.

Previously, I interned with the Vault🔐 Ecosystem team at HashiCorp and then in IBM 👁️🐝Ⓜ️'s Cloud division, building Terraform tooling and distributed CI systems for secure secret management. My efforts earned me the ✨Intern Communication Principle Award✨ at HashiCorp in 2024 and the ✨Intern Execution Principle Award✨ at IBM in 2025.

Long term, I want to help build ML systems that make state-of-the-art models accessible to teams and users far from big research labs.

Profile photo placeholder

Research Interests

  • Scalable and federated ML systems
  • Expert-routed / Mixture-of-Experts architectures
  • Optimization-guided model compression and structured pruning
  • Applications in NLP, Vision and decision-making systems (like financial markets)

Interests & Hobbies

  • 🚴‍♂️ Biking: recently started riding routes in the shape of countries (Tunisia and South Africa so far).
  • 🌍 Traveling: visited 30 U.S. states and 23 countries (and counting).
  • 🗣️ Learning languages: speak Arabic, English, French, and Spanish, and currently learning Italian.

News

  • 📢 Upcoming - I'll be attending NeurIPS 2025 in San Diego (Dec 2-8). Let's connect if you'll be there! See you in San Diego🌞🌴!
  • Oct 2025 - Released our new preprint TwIST: Rigging the Lottery in Transformers with Independent Subnetwork Training.
  • Sep 2025 - Our paper Learning to Specialize: Joint Gating-Expert Training for Adaptive MoEs in Decentralized Settings was accepted to NeurIPS 2025 🎉.
  • Sep 2025 - Accepted to the MENA ML 2026 Winter School (acceptance rate 13%)🎉 to present my work on federated Mixture-of-Experts for test summarization.
  • Aug 2025 - Concluded my software engineering internship in IBM's Cloud division (Vault / HashiCorp acquisition).
  • May 2025 - Graduated from Rice University with a B.S. in Computer Science🎉 and started my thesis-based M.S. in Computer Science at Rice.
  • May 2025 - Wrapped up my internship with the Vault Ecosystem team at HashiCorp.

Publications & Preprints

Peer-Reviewed Publications

  • Yehya Farhat, Hamza ElMokhtar Shili, Fangshuo Liao, Chen Dun, Mirian Del Carmen Hipolito Garcia, Guoqing Zheng, Ahmed Hassan Awadallah, Robert Sim, Dimitrios Dimitriadis, Anastasios Kyrillidis. Learning to Specialize: Joint Gating-Expert Training for Adaptive MoEs in Decentralized Settings. In Proceedings of NeurIPS, 2025. PDF · Poster · arXiv

Preprints / Under Review

  • Michael Menezes, Barbara Su, Xinze Feng, Yehya Farhat, Hamza ElMokhtar Shili, Anastasios Kyrillidis. TwIST: Rigging the Lottery in Transformers with Independent Subnetwork Training. Preprint (Under review), 2025. PDF · arXiv
  • Hamza ElMokhtar Shili, Natasha Patnaik, Isabelle Ruble, Kathryn Jarjoura, Daniel Suarez Aguirre. Projection-Free CNN Pruning via Frank-Wolfe with Momentum: Sparser Models with Less Pretraining. Preprint, 2025. PDF · arXiv

Posters & Presentations

  • Satellite Segmentation for In-Space Inspection Hamza ElMokhtar Shili, Sophie Chikhladze, Taeho Choe, Aditi Balaji, Jeffery Huang Rice Data to Knowledge (D2K) Spring 2025 Showcase, Rice University — Houston, TX — May 2025 Poster
  • Reinforcement Learning for Financial Market Optimization: Risk-Aware Policy Design Hamza ElMokhtar Shili, Andres Villada, Shahrazad Ali 11th Eubank Conference on Real World Markets, Rice University — Houston, TX — Apr 2024 Poster
  • Generating Low-Depth Quantum Variational Circuits for the Hydroxyl Cation to Attain High Fidelity on NISQ Computers Jason Han, Jason Ludmir, Hamza ElMokhtar Shili, Tirthak Patel Quantum Computing for Drug Discovery Challenge, ICCAD 2023 — San Francisco, CA — Oct 2023 Report · Code

Teaching & Service

Teaching

  • COMP 414/514: Optimization - Algorithms, Complexity & Approximations - Rice University - Head TA
  • COMP 412: Compilers Design and Construction - Rice University - TA
  • COMP 321: Intro to Computer Systems - Rice University - TA
  • COMP 382: Reasoning about Algorithms - Rice University - TA

Mentoring & Service

  • Mentor for students from my former high school in South Africa applying to CS programs; helped organize a hackathon to build their coding skills.
  • Mentor for undergraduate students interested in scalable and federated ML systems (MoE routing and pruning).
  • Volunteer for outreach workshops introducing high-school students in Africa to programming and AI through regional initiatives.

Contact

Email: hamza [dot] shili [at] rice [dot] edu

Institution: Rice University

Location: Houston, TX