Seattle’s OneClick.ai — Bringing AI a Click Closer

OneClick.ai, a startup founded by two former Microsoft engineers in Seattle, is on a mission to make AI more accessible to businesses. “We design, build and deploy custom AI models as a scalable API that can be accessed from anywhere. Just prepare your data, and we’ll take care of the rest,” says co-founder and CTO Ning Jiang.

The industry is in need of automated AI solutions: product and service development cycles can stretch up to six months, but experienced AI engineers and data scientists are costly and in short supply.

OneClick’s platform frees up busy AI engineers and data scientists from the model design pipeline by performing data cleansing, feature engineering, algorithm design, hyperparameter tuning, iterative modeling, and model assessment automatically within a matter of hours. It can solve regression, classification, time-series forecasting, clustering, recommendation, and vision problems for businesses in any industry or sector.

A client simply uploads their numeric, categorical, date/time, text, or image data to the platform, defines the task and OneClick’s AI software does the rest automatically. Its on-call chatbot “Eva” can assist customers with data analysis, model training, and evaluation questions. The finished model can be accessed from anywhere using a web API.

OneClick says their closest competitor today is Google Cloud AutoML (released in January 2018). AutoML is powered by a core technology called progressive neural architecture search (PNAS) — which supports scenarios around computer vision models. PNAS trains models two times faster than previous reinforcement learning methods, and five times faster than evolution algorithms.

Jiang explains, “PNAS algorithms are domain specific and support only neural networks, depending heavily on human heuristics. They are also very expensive to train, demanding thousands of GPU hours. There’s also a cold start problem: NAS has no prior knowledge about data and works based on assumptions.“

Another competitor is Microsoft Custom Vision Services, launched in 2017, which supports customized vision models similar to Google AutoML.

OneClick’s AI is powered by a technology called Generalized Architecture Search (GAS), which is capable of handling more complex data and predictive tasks. GAS supports both neural networks and traditional machine learning algorithms, has automated feature learning, and fewer models to train. The algorithm observes and applies domain knowledge learned through a supervised learning algorithm to the original data. Each custom model can be iteratively improved.

For example, in the image classification task tagging running horses or/and drinking horses, which uses a small sample size for training and testing, the OneClick model achieved over 95 percent accuracy in comparison to Google’s AutoML’s 85 percent and Microsoft Customer Vision’s 75 percent.

“The hardest part for us was tackling the technical challenge of automated machine learning,” recalls company co-founder and CEO Shen. The team began searching for investment in early 2017. Sinovation Ventures, a tech-savvy Chinese investment firm that manages USD and RMB investment funds and is led by well known AI investor Kai-Fu Lee, evaluated the company’s technical abilities for months before adding it to their investment portfolio.

Shen has over ten years of experience in AI across multiple industries. He was one of the first to leverage deep learning technologies to overhaul eBay’s 2C sales flow. He was also heavily engaged with the Microsoft Bing search engine, designing various AI models for image search and paid ads. Before working at Microsoft Shen was an early adopter in the application of AI technologies for tumor detection from colon CT images, and built an image-based system to guide doctors during surgical procedures.

Jiang meanwhile has over 15 years of varied AI experience. He was a senior manager at Microsoft responsible for Bing Maps Autosuggest, and managed Bing Local Search in Europe. He previously led development of the Bing Ads relevance platform and algorithm improvements, delivering paid search experiences in both Bing and Yahoo.

The duo said they decided to start their own company after witnessing “the increasing gap between the rapid development of AI technology and slowly-catching-up business applications.”

---

Journalist: Meghan Han | Editor: Michael Sarazen