Research Scientist — AI Systems & Trustworthy Machine Learning
Jamil Gafur
Assistant Research Scientist, Flatiron Institute / Polymathic AI · Ph.D. candidate, University of Iowa (Spring 2027)
I design and deploy resource-efficient, explainable AI systems — from agentic reasoning architectures to billion-parameter scientific foundation models. Six+ years across Flatiron Institute, Siemens, Cornell University, and the national laboratories (NREL, LANL).
What I Do
Agentic AI Systems
Distributed multi-agent orchestration and reasoning: recursive task decomposition into Directed Acyclic Graphs, persistent per-path knowledge blackboards, and MCP-compliant tooling across heterogeneous LLMs.
Foundation-Model Optimization
Production inference pipelines for large scientific foundation models — quantization, compression, and benchmarking automation that preserve statistical reliability while radically reducing compute and memory footprints.
Trustworthy Machine Learning
Explainable AI (XAI), adversarial robustness evaluation, and neural network pruning — identifying what models actually learn, stress-testing them against edge cases, and compressing them by up to 90% without sacrificing accuracy.
Selected Impact
Recent Research
- Exploring the Structure Resulting from Unstructured Neural Network Pruning. IJACSA 17(6), 2026
- Structured Layer Pruning through Activation Variance. CSCI’26 — accepted
- Adversarial Attack of Sequence-Free Enhancer Prediction Identifies Chromatin Architecture. Bioinformatics, 2025 · DOI: btaf371
Open to collaborations in efficient ML systems, agentic architectures, and explainable AI — academic or industrial. Get in touch