About
Research intern at UIUC Blender Lab working on energy-based models. MSc AI (Distinction) from Essex, supervised by Prof. Luca Citi.
Research Interests
My ultimate objective is to study the cognitive mechanisms underlying intelligence and develop embodied agents able to reason and interact with the real world.
I work on differentiable approaches for iterative reasoning and planning in uncertain environments. Building more capable, efficient, and generalizable AI systems, particularly at the intersection of deep learning, statistical mechanics, probabilistic modeling, and geometric methods.In a wider context, I am interested in developing a network of specialized networks such as memory and rapid learning, goal-driven planning, spatial reasoning, and error detection and conflict monitoring that operate collaboratively to consistently make good decisions within a complex environment.
▸ Key areas (click to expand)
- Generative Modeling: Energy-Based Models, Diffusion/Score-Based Models, Normalizing Flows, Optimal Transport, Consistency Models. Mathematical tools: ODEs, PDEs (Fokker-Planck), SDEs, Itô Calculus, OT theory. Also language models and autoregressive generation.
- Generalisation, Reasoning & Planning: Robust OOD performance for complex decision-making.
- Geometric & Mathematical Foundations: Differential geometry (Riemannian manifolds), metric learning, Hamiltonian & Lagrangian formulations for efficient learning algorithms.
- Efficient Architectures: Transformers & attention mechanisms.
- MPC, RL, Agents & Embodied Intelligence
- Applications: World Models, robotics, AI for science, LLMs.
Current Focus: Unifying diffusion, flow-based, and energy-based models through statistical mechanics and differential geometry, toward agents capable of fast generation, reasoning (as optimisation), and robust planning under uncertainty.
Research Highlights
- Reasoning & Planning: Agents that reason, plan, and adapt to novel environments
- Representation Learning: Learning the geometry of parameters and data distributions
- OOD Generalisation: Robust behaviour beyond training distribution
- Embodied Intelligence: Systems that learn through interaction with the world
Open Source
Current libraries
- TorchEBM 🍓 Flagship
· EBMs, diffusion, flow matching, SDE/ODE integrators
- cuRBLAS 🍒 · CUDA kernels for randomized linear algebra
▸ Previous libraries (click to expand)
- TransformerX · Modular transformer research library
- Emgraph · Knowledge graph embeddings
- Bigraph · Bipartite graph link prediction
- Nano Autodiff · From-scratch autodiff engine
- TASE · Large-scale multimodal audio search engine
News
- Started Research Collaboration with UIUC Blender Lab - Jul. 2025
- MSc defended at Essex (NIR thesis) - Oct. 2024
Awards
- Full scholarship for MSc AI, University of Essex
Education
University of Essex
MSc Artificial Intelligence (Distinction)
2023 - 2024
Dissertation: NIR — Neural Integration of Iterative Reasoning in LLMs for Code Generation
🥇 1st place, Inter-Departmental Neural Network Challenge (100+ participants)
University of Kurdistan
BEng Computer Engineering
2014 - 2018
Thesis: EfficientCoF — Subspace clustering for collaborative filtering
Experience
UIUC Blender Lab
Research Intern
Jul 2025 - Present
Energy-based transformers for image/video generation. Investigating mode collapse, inference-time compute scaling, and fast sampling strategies.
University of Essex
Postgraduate Researcher
Apr 2024 - Oct 2024
Developed NIR framework for integrating iterative reasoning into LLM hidden states without fine-tuning.
Open Source
Developer & Maintainer
2019 - Present
TorchEBM (8K+ downloads), cuRBLAS, TransformerX, Emgraph, Bigraph, TASE. See Projects section.
Projects
TorchEBM 🍓
Flagship
PyTorch library for energy-based models, diffusion, and flow matching. Implements samplers, score/flow/contrastive objectives, SDE/ODE integrators, interpolants, and mixed-precision training.
PyTorch · CUDA
cuRBLAS 🍒
GPU-accelerated randomized linear algebra. CUDA kernels for Hutchinson trace estimation, randomized SVD, and probabilistic matrix operations—useful for sliced score matching and large-scale ML.
C++ · CUDA · Python
TransformerX
Modular transformer research library. Composable attention mechanisms, positional encodings, and architecture variants for rapid prototyping.
TensorFlow
Emgraph
Knowledge graph embedding library. Train and evaluate TransE, DistMult, ComplEx, and other relational models on link prediction tasks.
TensorFlow 2
Bigraph
Bipartite graph link prediction. Implements Jaccard, Adamic-Adar, Common Neighbors, Preferential Attachment, and Katz similarity for two-mode networks.
NetworkX · NumPy
Large-scale audio search engine. Scalable indexing and retrieval with Elasticsearch, ArangoDB, and Redis for real-time multimodal search.
Elasticsearch · Python
