I shipped a production Risk AI Assistant and rebuilt a millions-row migration path around automated loading.
Daily backfill18 hours
Legacy path · manual bottlenecks
54× faster95–97% assistant accuracy8 weeks concept to production
University of Michigan · Cybersecurity Lab
Cybersecurity Research Assistant
Sep 2025 — Jan 2026
I tested how jailbreak attempts evolve across a conversation, then designed a similarity hash to surface iterative steps.
One prompt looks harmless in isolation. Session context reveals the sequence.
90%+ attacks iterative17 embeddings evaluated0.8–0.9 early recall
NJIT · ACAD Lab
Photonic & Software Engineering Researcher
May — Oct 2025 · Newark
I helped translate natural-language device goals into generated, simulated, and optimized photonic circuits.
Weeks of hand-tuning between layout, simulation, and verification.
80% higher functional yield68%+ pass@k gain40+ public designs
QCRI · aiXamine
Software Engineering Intern
May — Jul 2024 · Doha
I built a robustness evaluation system that made model behavior under distribution shift measurable.
Calibration error
9.4% → 3.1%
Pipeline time
3.2h → 28m
14 LLMs62 distribution shifts150k prompts+21% AUROC
Published research · July 31, 2026
When AI meets photonics.
PICasso: An AI-Enabled Design Framework for Autonomous Optimization of Silicon Photonic Devices
IEEE International Conference on LLM-Aided Design (ICLAD) · NSF REU research at NJIT
36 benchmark tasks92.7% structural satisfaction1.74 dB lower average loss
The framework connects natural-language specifications to YAML-to-GDS generation, automated placement and routing, verification, and simulation-guided optimization.
The foundation
Education
University of Michigan · Ann Arbor
B.S. in Computer Science
Computer Architecture and AI track · GPA 3.8 / 4.0 · Expected May 2027
Selected coursework
Computer Security · Advanced Operating Systems · Computer Organization · CS Pragmatics · Linear Algebra · Data Structures and Algorithms · Human-Centered Software Engineering
Tools of the trade
Technical skills
Machine learning
PyTorch, scikit-learn, NumPy, pandas, FastAPI, SciPy, Seaborn, Hugging Face Transformers