Adam Sabry

WORK HISTORY

EDUCATION

ABOUT ADAM SABRY

I build machine learning systems that operate at production scale.At NASA Ames Research Center, I train deep learning models on multi-terabyte scientific datasets across multi-GPU HPC clusters, engineer RAG-based AI workflows as part of the NASA GPT initiative, and optimize distributed training pipelines for real-world research impact.My focus areas:Time Series Deep Learning & Forecasting:Trained Mamba-based models for seismic ground-motion forecasting on terabytes of continuous seismic data (Japan & California), co-authored AGU 2025. Built a Temporal Fusion Transformer for NBA player performance prediction with 77% directional accuracy.Machine Learning Systems & Deployment:Built end-to-end ML systems from training to production: RAG-based NLP pipelines (NASA GPT), computer vision systems for scientific plot understanding (85–90% accuracy across 5+ chart types), and interpretable time series models. Deployed models on Hugging Face with Gradio UIs for real-world use.Distributed Training & GPU Optimization:Optimized multi-GPU training pipelines (PyTorch DDP, NCCL) improving GPU utilization by ~20–30%. Debugged distributed training bottlenecks including NCCL communication issues, topology mismatches, and memory constraints. Profiled workloads with Nsight Systems/Compute and nvidia-smi to classify bottlenecks as compute/memory/latency-bound.Projects & Publications:• Multi-domain AI life recommender trained on 28M+ ratings across movies, food, and books. Live on HF.• NBA Player Performance Forecast: TFT model predicting player improvement/regression with explainable AI. 77% directional accuracy. Live on Hugging Face.• Scientific Plot Understanding: Computer vision pipeline extracting structured data from scientific charts via OCR and segmentation. 85–90% accuracy across 5+ chart types.• Seismic Ground Motion Forecasting: Co-author, AGU 2025 – applying Mamba models to earthquake forecasting on multi-terabyte datasets.• NASA GPT: RAG-based LLM system for automated support workflows using vector search and semantic context injection.Background:Currently pursuing MS in Artificial Intelligence at San Jose State (expected May 2027, 4.0 GPA). Experienced leading teams of 25+ in high-pressure environments (General Manager, Five Guys).I\'m seeking roles in:Machine Learning Engineer | Deep Learning Engineer | ML Infrastructure Engineer | Solutions Architect | Data Scientist | LLM/GenAI EngineerWhether you\'re building AI products, training at scale, or optimizing ML infrastructure let\'s connect. Email: A••••••••@gmail.com