Joel Hestness

Joel Hestness is a computer-systems and machine-learning researcher who co-founded 3 Day Startup, the university-focused entrepreneurship program from which his investor contact ([hidden]) originates. 3 Day Startup was launched in 2008 at the University of Texas at Austin by Bart Bohn, Joel Hestness, and Thomas Finsterbusch, originally as a student organization. It has since become a 501(c)(3) non-profit that runs intensive, hands-on weekend programs teaching students how to evaluate ideas, build teams, and launch companies, expanding to schools including Harvard and MIT and to campuses across North and South America, Asia, Africa, and Europe.

Academically, Hestness studied at the University of Texas at Austin (roughly 2007–2011) before earning his Ph.D. in high-performance computer architecture from the University of Wisconsin–Madison (2012–2016). During his studies he interned in chip design and hardware-security research (including stints associated with AMD, Intel, and Microsoft Research) and earned an IEEE Micro "Top Picks" recognition for his computer-architecture work, which he presented at venues such as the International Symposium on Computer Architecture (ISCA).

Hestness subsequently moved into deep-learning research. At Baidu's Silicon Valley AI Lab (SVAIL) he was among the first to demonstrate predictable, empirical accuracy scaling laws for modern deep-learning models — foundational work that helped spark the now-pervasive field of LLM scaling-law research. He has also held research roles at AMD Research and NVIDIA Research.

Today he is a Distinguished/Principal Research Scientist and Core Machine Learning Team Lead at Cerebras Systems, the maker of wafer-scale AI accelerators. His work there has produced compute-efficient scaling laws (Cerebras-GPT) and contributed to state-of-the-art open models including BTLM, Jais, and CrystalCoder, focusing on natural-language applications and their scaling, training dynamics, and efficiency.

• Operates primarily through 3 Day Startup, which sources, mentors, and helps launch university and student-founded ventures at the idea/pre-seed stage rather than running a conventional fund.
• Stage emphasis is pre-seed and seed — backing founders at the earliest, team-and-idea-formation phase, consistent with 3DS's "go from idea to launch in a weekend" model.
• Given his research career, his domain credibility and likely interest areas center on AI/ML, deep learning, large language models, and computing/semiconductor hardware.
• A "yes" is plausibly anchored in 3DS values: committed founders, fast execution, hands-on validation, and a willingness to test ideas with real users — the core behaviors the 3 Day Startup program is built to reward.
• Note: no public, itemized personal angel portfolio (specific company-by-company deals or check sizes) could be verified from primary sources; his investor footprint is tied to the 3 Day Startup organization.