This article recounts the entrepreneurial journey of TQ (灶爷), CEO of Huhu AI, from his early days tinkering with discarded computers to leading a company acquired by Amazon. It highlights the unique landscape of the early AI era, where acquisition was often the primary path to profitability.
Early Beginnings and Resourcefulness
Building a Lab from Scraps
TQ's journey began with a resourceful approach to building a computer lab. He and his partner collected discarded computers from university dumpsters, piecing together approximately six machines. This makeshift lab, housed in their dorm room, was powerful enough to frequently overload the electrical circuits.
Early Exposure to Computer Vision
TQ's initial foray into computer vision occurred around 2007 at Harbin Institute of Technology (哈工大). At the time, the field was largely focused on methods like SVM, and neural networks were a niche topic. This early exposure, however, laid the foundation for his future endeavors. He noticed that many graduates from his lab primarily pursued research careers, highlighting a lack of commercial opportunities in the field.
Navigating the Early AI Landscape
The Influence of Early AI Research
TQ attended a lecture by Fei-Fei Li at MIT in 2011, where the accuracy rates of ImageNet competitions were around 40%. The emergence of AlexNet in 2012 marked a turning point, although its initial accuracy of 64% was still far from production-ready. These early advancements, while promising, underscored the limitations of AI technology at the time.
A Serendipitous Start
The decision to start his company wasn't driven by foresight about AlexNet's impact. Instead, TQ found his doctoral research at Boston University, which involved tracking bat migrations in Texas, uninspiring. This dissatisfaction led him to explore entrepreneurial pursuits in computer vision.
Building a Cluster and the Birth of Orbeus
While looking for jobs, TQ and his partner built a computer cluster from discarded parts. They then sought to utilize then nascent open-source cloud computing projects and apply them to computer vision algorithms. This was enabled by an increase in open-source big data processing technologies, such as Hadoop, and the relatively open-source nature of the internet giants like Facebook and Google. That initiative ultimately evolved into Orbeus, a computer vision API company. Orbeus offered a simple web interface, reminiscent of Google, where users could upload images and receive information about their content.
From Startup to Acquisition
Incubator Success and Full-Time Commitment
Orbeus gained traction, participating in entrepreneurial competitions and being accepted into the Techstars incubator program. This prompted TQ and his co-founders to make the decision to commit fully to the startup. They were the first ethnic-chinese project to be accepted to Techstars. The decision was made easier by the fact that software engineering was relatively lenient as a career at the time, making the opportunity to join a top incubator too good to pass up.
A "Hedgehog" Approach to Innovation
When faced with the decision of focus vs diversification, TQ compared his team to a hedgehog that knew one thing well and stuck to it. This resonated with his team, who were deeply familiar with computer vision and preferred to focus on their area of expertise. This strategy ensured that they remained specialized.
The Unique AI Landscape of the Early 2010s
The early 2010s was a unique period for AI startups. Investment was aggressive due to low interest rates, but AI technology was not yet fully mature. As a result, many companies, including Orbeus, found that their primary path to profitability was through acquisition rather than direct sales.
Acquisition by Amazon and Beyond
Orbeus was acquired by Amazon for approximately $50 million after raising around $5 million in funding. This trend of acquisitions was common at the time, with larger companies like Google, Facebook, and Intel acquiring AI startups for their talent and expertise. TQ and his team then became the core of what became the now widely known Rekognition visual AI services on AWS. After three and a half years at Amazon, he joined Wyze, a Seattle-based camera startup, as CTO, building their technology team from scratch to over 150 people. TQ is now CEO of Huhu AI, a company focused on AI solutions for the fashion industry.
{# 
{# 
{# 
{# 
{#