Profile banner
Vahid Kazemi

Vahid Kazemi

Machine Learning PhD, Staff Engineer at OpenAI, ex-Apple, Google, Waymo

San Francisco, California

Summary

Vahid Kazemi is a distinguished Machine Learning expert with a Ph.D. from KTH Royal Institute of Technology, specializing in computer vision, natural language processing, and robotics. His academic contributions include influential papers like 'One Millisecond Face Alignment with an Ensemble of Regression Trees' (CVPR 2014), which is highly cited, and 'Show, Ask, Attend, and Answer: A Strong Baseline For Visual Question Answering'. His work has also led to several patents in image composition and enhancement using machine learning. vahidkazemi+3
With a rich career trajectory, Vahid has held significant roles at leading technology companies. He is currently a Member of Technical Staff at OpenAI, having previously served as a Senior Machine Learning Researcher at Apple. His experience also includes Engineering Manager positions at Pinterest and Snap Inc., where he led teams in visual search and discovery, and a Software Engineer role at Google, where he contributed to Waymo projects and pioneered best practices for TensorFlow and PyTorch. vahidkazemi+3
Vahid actively shares his insights and expertise in AI and software engineering. He maintains a personal website and GitHub repositories for 'EffectiveTensorFlow' and 'EffectivePyTorch', offering guides and best practices. He is also an active presence on X (formerly Twitter), where he frequently tweets about developments in AI, LLMs, and software architecture. Additionally, he creates video content on YouTube, discussing machine learning and technology topics. vahidkazemi+5
Vahid demonstrates a keen interest in the philosophical and practical aspects of Artificial General Intelligence (AGI) and large language models (LLMs). He has publicly stated his opinion that AGI has already been achieved, often engaging in discussions about the capabilities of neural networks, the nature of creativity in AI, and the evolution of AI model training paradigms, advocating for simpler, scalable solutions and continuous model improvement. nitter
A proponent of robust software engineering principles, Vahid has expressed views on modularity and practicality. He advocates for the use of small, low-dependency libraries over monorepos and emphasizes starting with specific, minimal implementations that evolve into generalized solutions as real use cases emerge, rather than prematurely designing for imaginary broad applications. nitter

Work

Education

Projects

Hobbies

Creates video content on YouTube about machine learning and technology. vahidkazemi+1

Enjoys movies, particularly strong narratives and creative direction. nitter