“Before Kevin led the ML Specialist team, where I worked with him, he was on the Data Analytics team, giving him an unusual breadth in the Data Science field. On our team, I saw the depth of his skills: streamlining a complex CV model process from months to days, coaching an NLP customer on the best batch management to utilize hardware accelerators, writing a TF code tutorial for the global specialist team, and so much more. He’s the endlessly curious kind of person who picks up Kubeflow over the weekend. But, while it is unusual to have experience across the full pipeline, it is a truly rare gift to combine deep technical skills and intense drive with humility and emotional intelligence. Kevin is that rare person.”
Activity
-
Off to Neurips '24! Looking forward to catching up with old friends and meeting some new faces. Google always has such an inspiring presence at this…
Off to Neurips '24! Looking forward to catching up with old friends and meeting some new faces. Google always has such an inspiring presence at this…
Liked by Kevin Tsai
-
Dear Google Team, As I prepare to conclude my six-year journey with Google Cloud, I want to express my heartfelt gratitude to each of you…
Dear Google Team, As I prepare to conclude my six-year journey with Google Cloud, I want to express my heartfelt gratitude to each of you…
Liked by Kevin Tsai
Publications
-
Testing Image Understanding through Question-Answering
In recent years, neural networks have produced some very exciting results in diverse computer vision tasks, such as image recognition, image captioning, artistic style transfer, auto-coloring, segmentation and image generation. In this paper we ask a fundamental question: ”How well do these networks really understand images?” We explore this question through through two approaches: first through the task of Visual Question Answering (VQA), where a model is trained on images and associated…
In recent years, neural networks have produced some very exciting results in diverse computer vision tasks, such as image recognition, image captioning, artistic style transfer, auto-coloring, segmentation and image generation. In this paper we ask a fundamental question: ”How well do these networks really understand images?” We explore this question through through two approaches: first through the task of Visual Question Answering (VQA), where a model is trained on images and associated question answer pairs in natural language. Next, we measure effectiveness of transfer-learning to image recognition. We use a model pretrained on image-recognition task and retrain its weights during VQA training and then test it again on image recognition task.
Other authors -
-
Peer Lending Risk Predictor
Abstract - Warren Buffett famously stated two rules for investing: Rule #1. Never lose money, and Rule #2. Never forget Rule #1. Recent Peer Lending opportunities provide the individual investor the opportunity to earn an interest rate significantly higher than that of a savings account. However, a default on the loan by the borrower means the investor will lose her entire principal. In this paper, we will use Machine Learning algorithms to classify and optimize peer lending risk. We will also…
Abstract - Warren Buffett famously stated two rules for investing: Rule #1. Never lose money, and Rule #2. Never forget Rule #1. Recent Peer Lending opportunities provide the individual investor the opportunity to earn an interest rate significantly higher than that of a savings account. However, a default on the loan by the borrower means the investor will lose her entire principal. In this paper, we will use Machine Learning algorithms to classify and optimize peer lending risk. We will also demonstrate how our classifier can identify special gems. These are outlier loans with unusually high interest rates, at up to 18%, doubling the rates offered by peer lending sites classification for the in the same risk category.
Other authors -
Patents
-
DATABASE MANAGEMENT SYSTEM RISK ASSESSMENT
Issued US US Patent 8,676,746
See patentA method of evaluating an implementation of a DBMS is provided. The method comprises collecting data associated with the implementation of the DBMS and accessing a database comprising problems and their associated solutions, wherein the solutions are configured to remedy at least one of the problems. The method further comprises comparing the data associated with the implementation of the DBMS with the problems and identifying at least one problem associated with the DBMS. Finally, a DBMS risk…
A method of evaluating an implementation of a DBMS is provided. The method comprises collecting data associated with the implementation of the DBMS and accessing a database comprising problems and their associated solutions, wherein the solutions are configured to remedy at least one of the problems. The method further comprises comparing the data associated with the implementation of the DBMS with the problems and identifying at least one problem associated with the DBMS. Finally, a DBMS risk assessment report is generated that identifies the problem associated with the DBMS and a solution configured to remedy the problem.
Recommendations received
7 people have recommended Kevin
Join now to viewMore activity by Kevin
-
I bit of a late post, but October 1st was my last day at Google. I want to thank the Geo Leadership Team for the opportunity to have built the first…
I bit of a late post, but October 1st was my last day at Google. I want to thank the Geo Leadership Team for the opportunity to have built the first…
Liked by Kevin Tsai
-
Over the past three weeks, I have traveled to multiple cities Dallas, NYC and Boston presenting AI Agents and Evaluation related topics. We had…
Over the past three weeks, I have traveled to multiple cities Dallas, NYC and Boston presenting AI Agents and Evaluation related topics. We had…
Liked by Kevin Tsai
-
As we create more AI application, Kubernetes has become necessary part of the infrastructure to scale AI model training and deployment. The demand…
As we create more AI application, Kubernetes has become necessary part of the infrastructure to scale AI model training and deployment. The demand…
Liked by Kevin Tsai
-
I'm excited to announce my return to the Games Industry as the Go To Market Leader for Games here at Databricks. Like 37% of gamers out there it is…
I'm excited to announce my return to the Games Industry as the Go To Market Leader for Games here at Databricks. Like 37% of gamers out there it is…
Liked by Kevin Tsai
-
⏰ Countdown started! 🚀 In few days, Holt S. and I are going to present a step-by-step guide on building a multimodal Retrieval Augmented…
⏰ Countdown started! 🚀 In few days, Holt S. and I are going to present a step-by-step guide on building a multimodal Retrieval Augmented…
Liked by Kevin Tsai
Other similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Kevin Tsai in United States
-
Kevin Tsai
-
Kevin Tsai
Associate at Brookfield Asset Management
-
Kevin Tsai
-
Kevin Tsai
Mathematician specialized in designing and applying mathematical model to complex systems
63 others named Kevin Tsai in United States are on LinkedIn
See others named Kevin TsaiAdd new skills with these courses
-
13h 18m
VMware vSphere 8 Certified Professional - Data Center Virtualization (VCP-DCV) (2V0-21.23) Cert Prep
-
12h 48m
VMware vSphere 8 Certified Technical Associate - Data Center Virtualization (VCTA-DCV) (1V0-21.20) Cert Prep
-
1h 42m
Cloud Computing Careers and Certifications: First Steps