Welcome

I am a Machine Learning Engineer with around 4 years of experience specializing in developing and deploying machine learning solutions. At Axiado Corporation (Feb 2025 – Present, San Jose, CA), I design and deploy machine learning models on edge devices using TensorFlow Lite for efficient real-time inference. My work includes implementing quantization-aware training (QAT) and post-training quantization (PTQ) to optimize models for edge deployment, as well as performing model profiling and performance optimization to ensure low-latency and resource-efficient execution. Previously, at Goldman Sachs, I created anomaly detection systems that prevented monthly write-offs of $1-2 million, demonstrating my ability to deliver impactful solutions.

Having completed my Master’s in Computer Science at the University of California, Riverside in December 2024 🎓, I am eager to apply my advanced ML techniques to solve real-world problems, and I’m open to collaborations and opportunities.

Python C++ SQL JavaScript PyTorch TensorFlow Lite torch.compile Triton CUDA Scikit-learn NumPy Pandas Model Quantization (QAT/PTQ) Edge Deployment Performance Optimization Model Profiling Knowledge Distillation LoRA Fine-tuning Large Language Models (LLM) Deep Learning Anomaly Detection Computer Vision Recommendation Systems Reinforcement Learning AWS Docker Kubernetes MLOps

News and Updates

	Feb 2025 - Joined Axiado Corporation as a Machine Learning Engineer (Full-time) in Feb 2025 (San Jose, CA, On-site). Designed and deployed machine learning models on edge devices using TensorFlow Lite for efficient real-time inference. Implemented quantization-aware training (QAT) and post-training quantization (PTQ) to optimize models for edge deployment. Performed model profiling and performance optimization to ensure low-latency and resource-efficient execution.
	Sep 2023 - Completed Master's in Computer Science in December 2024, enhancing my industry experience with advanced theoretical and practical knowledge.
	Feb 2023 - Joined Bluevoir Technologies as a Principal Engineer, where I fine-tuned domain-specific LLMs to automate clinical trial protocol generation, reducing creation time by 30% and improving protocol accuracy by 40%.
	Aug 2021 - Joined Goldman Sachs as an analysts, where I developed an anomaly detection system that prevented $1-2 million in monthly losses by identifying fraudulent Apple Card transactions, and automated the dispute case process, reducing processing time by 40%.
	July 2019 - Joined Pegasystems in July 2019. Addressed critical issues, preventing major downtime for key clients and avoiding significant revenue losses.

Recent Blogs

Oct 2, 2025ML4LM — Speculative Decoding — from where we left off

Most blogs stop at the basics and skip the real details. I break down what's usually missing: batching, accept/reject checks, and fallbacks.

Sep 11, 2025ML4LM — Profiling torch.compile on DenseNet-121 Inference (GTX 1650)

Deep dive into profiling torch.compile performance on DenseNet-121 inference using GTX 1650, exploring optimization techniques and performance metrics.

Aug 24, 2025ML4LM — Guards vs Graph Breaks in PyTorch: What You Need to Know

Understanding the differences between guards and graph breaks in PyTorch's compilation system and their impact on model performance.

Aug 23, 2025ML4LM — A tiny Triton primer (toy example)

A beginner-friendly introduction to Triton with practical examples and toy implementations to get started with GPU kernel programming.

Aug 17, 2025ML4LM — PyTorch — What Not to Do in PyTorch Models for Better Performance (dynamo)

Common pitfalls and anti-patterns in PyTorch model development that can hurt performance, especially with Dynamo compilation.

Jan 24, 2025Knowledge Distillation at a Low Level

Exploring knowledge distillation techniques from a low-level implementation perspective, understanding how knowledge transfer works under the hood.

click here for more

Hoyath Ali