The Entrepreneurial Journey of a Poor Student From Peking University and Tsinghua University: The New Era of AI Has Arrived, and You Would Be Sorry If You Don’t Participate

Contents at a glance:Recently, HyperAI had the honor of meeting the OAISIS team, the second prize winner of the SegmentFault AIGC Hackathon 2023 Beijing, and talked with them about their mental journey in this competition and the work and life of the team outside the competition. During the conversation, the confidence, composure, and calmness displayed by the three young people were impressive.

Keywords:AIGC OAISIS Team Large Model Deployment   

This article was first published on HyperAI WeChat public platform~

Hackathon is a technology activity originated from Silicon Valley in the United States. Its specific method is to gather a group of developers to develop a work within dozens of hours. During the event, developers complete their ideas in the shortest time, showing their creativity and passion.It can be called "the coolest developer carnival in the world".

Recently, HyperAI participated in SegmentFault AIGC Hackathon 2023 BeijingDuring the competition, I personally experienced an exclusive party about AIGC.HyperAI was fortunate to have contact with the OAISIS team, which won the second prize in this competition.We talked with them about the team's feelings about the competition and the team members' work and life. During the more than two-hour conversation, HyperAI deeply felt the drive and enthusiasm of these young people in the face of opportunities, and was also attracted by their confidence and relaxation.

OASIS award-winning group photo

Self-proclaimed poor students: the path of growth for students from Peking University and Tsinghua University

The three members of the OAISIS team all come from top universities in the country. Team leader Eric Lu and team member Bai Yuan are alumni of Tsinghua University, and team member He Xinran is from Peking University.These three "children in the eyes of others" actually call themselves "bad students".To elaborate on the reasons, the three recalled their study experiences at Peking University and Tsinghua University.

Although the three of them entered university in different ways - Eric and Xinran took the college entrance examination, and Bai Yuanze was recommended to enter university by winning a competition, they all faced a huge adjustment in their mentality when facing university life. Eric said:"Calling myself a poor student is not self-deprecating. In fact, several of us have failed courses before."He and Bai Yuan are still upset about failing math in their first semester of college.

For three young people who had been "cheating" from elementary school to high school, the first taste of failure was not an easy thing to accept. In Bai Yuan's words, in order to digest this matter, he ran to the lotus pond of Tsinghua University and stared blankly for an entire afternoon. However,It is precisely such difficulties and setbacks that become a turning point in their mentality.

Being in a top domestic university that gathers the best talents from all provinces and cities, it is important to catch up with failure, but on the other hand, we also need to realize that there is more to it than the test scores.The more important significance of university is that it provides a platform that is free, inclusive, full of opportunities and challenges. Eric talked about how, through extensive exposure to cultural knowledge in college, communicating with teachers and classmates, or choosing to engage in scientific research, or starting to immerse himself in practice, he gradually found the direction in which he was good at and devoted himself to it, laying the foundation for his future career.

Now, the classmates have achieved outstanding results in their respective fields. The classmates who helped each other in their studies have now grown into helpers who rely on each other in the workplace."The connections we build during college have become a valuable asset in our lives."Bai Yuan said.

After graduating from undergraduate studies, the three chose to go abroad for further study. In the postgraduate stage, they no longer considered simply absorbing knowledge.Instead, we focus more on exploring new technical methods.At the same time, in the integration of different cultures and knowledge systems at home and abroad, the three people have gradually expanded their vision and thinking mode.

After completing their postgraduate studies, the three members officially entered the workplace. Eric joined ThomsonReuters, Bloomberg and other companies, and Bai Yuan worked as a senior engineer at Google.Three joined Snap lnc.It is a "camera company that is reinventing the camera to improve the way people live and communicate." The company's flagship product is a mobile application, Snapchat, that allows users to create and share Snaps, and its main source of revenue is digital advertising business.

According to Eric, OAISIS members are part of the SnapChat content platform team.Eric is responsible for coordinating the entire middle office department, and Bai Yuan serves as the technical director.The specific responsibilities of the middle office department are to provide underlying infrastructure support for SnapChat's three product lines - Friends Story, Discover Feed and Spotlight, involving data support, training limits of recommendation models, data indexing, and implementation of recommendation systems.Another team member, He Xinran, is mainly responsible for algorithm work.Iterate algorithms based on the middle platform to support product growth.

Eric: In this exciting era, you will be sorry if you don’t participate

The work experience at Snap allowed the three members to accumulate more understanding of the product and the overall construction process, and to have a complete grasp of the product at a certain level.They caught up with this big model wave,At the same time, we keenly captured the opportunities therein.

In the past six months, as the global big model competition has become increasingly fierce, LLM has been widely believed to have a structural impact on various industries. Many companies are in urgent need of quickly breaking through the high threshold of the basic layer of big models, namely computing power, storage and data, so that big models can quickly and efficiently promote business development. In this context,How to make large models produce truly effective transformation has become one of the key issues that technology practitioners are focusing on today.

Focusing on the three members, they noticed that due to the lack of large model development tools, the technical threshold for using AI large models is still high, and the ability to develop personalized applications is still in the hands of only a few teams with AI expertise, which is a big challenge for smaller companies with limited resources.They started to develop a low-code large-model application development tool.Reduce the manpower and time costs of enterprise application development, allowing users to focus on developing AI applications and business logic.

It was at this time that Eric received news about the AIGC Hackathon by chance. "Around mid-to-late March, a friend introduced us to this event." At first, they had no great desire to participate, but after serious consideration, the members agreed that perhaps through such an offline meeting, they would have the opportunity to meet more aspiring people in the AI circle. In the end,The decision was made to formally assemble and establish today’s OAISIS team.

In this Hackathon, OAISIS precipitated their thinking on the above challenges into a generative AI large model platform, and provided one-click deployment, testing, management, fine-tuning and other functions. It also supports the serial and parallel capabilities of different models, calling multiple models to jointly solve practical scenarios. Bai Yuan introduced: "We have innovations in technical details such as model deployment, and strive to provide some potential value for users." At the same time, he elaborated that for companies in the AI era, once they decide to launch a product, speed is the only way to win.The original intention of this product is to help users seize the market faster and compete for traffic dividends.

At the same time, according to the team, after this game,This AI big model platform has successfully attracted the attention of multiple investors.The team is currently actively contacting them.

Lu Qi recently said in a speech about the big model:"This era (the era of large models) is very similar to the gold rush era. If you went to California to pan for gold at that time, a lot of people would die. But people who sell spoons and shovels can always make money."Indeed, in the era of big models, it seems easier to do something in the application layer, but it is also difficult to do something different. How to build barriers, how to find differentiation, how to obtain financing, all these problems are facing entrepreneurs, and the OAISIS team is no exception.

However, in the face of these problems, the members showed a calm and optimistic side. The word Oasis is translated into Chinese as Green Oasis. When talking about this name, the team said that they chose this word because they compared the current large-scale model track to a desert.Every AI practitioner is exploring a way out, just like looking for an oasis.At the same time, they sincerely hope that the oasis can be found as soon as possible, ushering in a new era for AI practitioners.

With this expectation, the team chose to join this wave of entrepreneurship at this time. Although there are many like-minded people nowadays, the team believes that in such an era, the dreams and creations of every practitioner should be tolerated and encouraged.“We feel that this is a very exciting era and we would be sorry not to participate.” Eric said.

Confidence and relaxation should be the basic color of life

During the two and a half hour discussion,The three team members repeatedly mentioned several key words - confidence, choice and interest.Throughout their experiences, they always regarded these words as the mantra of life, and when faced with setbacks and pressure, they were able to maintain an objective self-evaluation and firmly choose the right direction at every stage of their lives.

He Xinran's personal experience is that his first choice in the college entrance examination was actually the School of Physics at Peking University. After entering the School of Computer Science, he also thought about transferring to the School of Physics in the second semester of his freshman year, but he was eventually "tortured" by his classmates in the School of Physics. Similar setbacks are inevitable in life.But He Xinran believes that we need to learn to reconcile with ourselves in the process.This kind of reconciliation is based on a full understanding of oneself - everyone has strengths and weaknesses, and ensures that one does everything with all one's strengths. As for the results, one must face them with a Buddhist attitude and analyze them rationally.

With this similar mentality, the three members met on campus, worked together in the workplace, and now choose to start a business together.We have maintained a good relationship all along."When facing difficulties, we can always remain calm and analyze and think, and at the same time communicate with each other honestly. I hope we can continue to work together in the future," Bai Yuan said with emotion when talking about team relationships.

This inevitably reminds me of what Wang Qiang, the founder of Zhen Fund, once said in an interview: after meeting 100,000 entrepreneurs,He believes that an excellent entrepreneurial team needs to have three qualities:

*  Especially pure,Full of passion for the field you enter.

*  The team is relatively complete.There should be at least two or three people who are willing to spend their youth and fight for something that seems like a dream with almost no salary.

*  Full of curiosity and ability to learn.Faced with the volatility of the market, we must have the courage to make transformations, even complete transformations, at certain stages.

In the OAISIS team, we see these three characteristics, as well as their confidence, calmness, and relaxation. Currently, the team is working hard to build its own products and has told HyperAI that the main battlefield in the future will be overseas. In addition,They also hope to recruit more like-minded friends to join the team.Let’s work together to find an oasis in our hearts. “Recruiting teammates mainly depends on fate. We believe in fate.”

At the end of the interview, HyperAI asked the team members what they would like to say to the future generations.The three summarized their experiences:

First of all, in the current environment of information explosion, I hope that students can maintain their own opinions and focus on areas of their interest.

Second, have a sense of ownership, approach problems with an all-out attitude, and be willing to take responsibility.

Third, in the face of new paradigms and new changes, we must assess the situation and seize opportunities.

Fourth, when facing pressure and anxiety, we must learn to adapt to the environment and resolve confusion.

In the future, HyperAI will continue to pay attention to the entrepreneurial journey of the OAISIS team.I hope their experiences can inspire more AI practitioners.

Guest introduction:

Eric Lu Cheng: Founder & CEO of OAISIS

Graduated from Tsinghua University and University of California, Berkeley, with a master's degree in computer science. Former head of Snap Inc. Content Infra, led the team to develop and operate a number of content social products based on Machine Learning, and worked as a senior engineer in companies such as Bloomberg and ThomsonReuters. Expert in cloud-native/multi-region service-oriented infrastructure, machine learning and recommendation system architecture, large-scale low-latency distributed storage solutions, and Internet-scale big data processing systems.

Yuan Baiyuan: Co-Founder & CTO of OAISIS

Graduated from Tsinghua University and the University of Washington, he holds a master's degree in electrical engineering and computer science. He was the technical leader of Snap Inc. Content Infra, where he developed and designed a large-scale content recommendation platform based on Machine Learning. He was a senior engineer at Google. He has extensive experience in large-scale distributed recommendation systems, big data processing systems, machine learning infrastructure, business logic abstraction, and API interface design.

Xinran He: OAISIS Co-Founder & Chief Scientist

Graduated from Peking University and University of Southern California, with a PhD in Computer Science, he has published a large number of academic results in multiple ML top conferences such as NeurIPS, ICML, KDD and received 2500+ citations. He was the leader of the push algorithm team at TikTok, responsible for the recommendation system and optimization of the entire Push & Notification chain. He served as the Content Ranking Tech Lead at Snapchat, responsible for the development and iteration of recommendation systems for multiple product lines including Friend Story, Spotlight, Growth, etc. As the author of the widely used recommendation algorithms (GBDT+LR, Deep & Wide) in the industry, he has a lot of expertise and experience in end-to-end large-scale machine learning solutions in the fields of Feed Ranking, Growth Ranking, Ads Ranking and Push Notification.

OASIS product website:https://www.oaisis.io/

This article was first published on HyperAI WeChat public platform~

-- over--