== Neural Market Trends ==

2020 AI Market in Review

AI Machine Learning

2020! What a crazy year! We had the Covid19 pandemic, major economic shocks, two big IPO’s in the immediate ‘AI’ space, and lots of new machine learning libraries and innovation.

Right now there are so many startups in so many different niche markets or services that it’s hard to count. I’ll highlight the ones that I stumbled across.

First is Spell.ml. A buddy of mine ended up there and clued me into what they do. They help manage, spin up, and allocate machine learning instances via a Python library. They seem to operate in the ML Ops part of ML pipeline as well and remind me a lot of H2O’s Puddle product. Managing your cloud instances is something that’s sorely needed, especially if you ever left a large instance ‘on’ and had to pay big money.

PyTorch had a big year, many of my colleagues like it better than Keras and TensorFlow and I see quite a bit of adoption happening. It seems that every 6 months or so there’s a major breakthrough in the deep learning research area and I look forward to more advances next year.

RapidsAI made a big splash this year and for good reasons. They’re GPU based libraries let’s you do ETL and other dataframe manipulations right on a GPU. Scaling on GPUs is getting easier. I’ve only played with RapidsAI and tried to install it on my Jetson Nano. It didn’t work because of the different architectures but I expect great things out of NVDA and RapidsAI next year.

A lot of news dropped during AWS Reinvent this December. I’m floored by the offerings they’re bring out to their customers. Sagemaker capabilities keep growing and they have a good story for a complete end to end pipeline for all machine learning and AI projects. Of course, you’re locked into their ecosystem! However, AWS is has a good siren call and many companies can’t ignore it any longer

Not to leave Azure and GCP out of the loop, all three cloud providers made big strides this year in their offerings. Before ReInvent 2020, I’d say that Microsoft’s Azure would win the cloud wars, now I’m not to sure. I do know that unless Google steps up, it will always be the third wheel.

My old friends at RapidMiner had a busy year, bringing new products to market. I believe their Turbo Prep product for ETL and AutoModel were released in 2019 but they started to pull together their ecosystem tighter with AI Hub. The neat thing they released was Jupyter Notebook integration. That’s a big plus in my mind. Still, I wish they would uncripple RapidMiner Studio and just make it completely open source. Maybe in 2021?

KNIME is growing too. In my chats with customers and prospects, KNIME shows up in different parts of an organization. Data Scientists, DevOps, and even the undefined Citizen Data Scientist seems to like their “no to low code” platform. To augment their “AutoML” capabilities, they recently partnered with H2O.ai by integrating the Driverless AI product.

H2O.ai had a great year. I do work there and what I’m writing here are my own viewpoints. Both main products, Driverless AI and the open source H2O-3 keep growing with new capabilities and features. H2O’s ML Ops product was released as well as the new open source Wave AI App development SDK. Wave is pretty cool, it has a GoLang backend for scalability and a Python 3.7+ framework to build AI Apps. You should check out the GitHub repo and give it a try.

It was a year or two ago that I heard of C3.ai and they just IPO’d a week ago. Just like the Snowflake IPO, it shot up a like a rock. Investors, Traders, Speculators, and everyone’s Uncle wants a piece of the AI market. C3.ai has some interesting products that connect to different open and closed source technologies and it will be interesting to see where they go in 2021. I wonder if Tom Siebel can pull off another Siebel Systems.

I mentioned Snowflake above and they had a great IPO too. Not quite in the AI space, they partnered with H2O.ai and other AI companies to start embedding AI technology into their software. H2O.ai did that with their MOJO’s and Driverless AI. Anyone that knows SQL can call a trained model or train a model right within their interface.

2020 also saw year where more venture capital flooded the market. Dataiku, Datarobot, and Streamlit got raises and many smaller companies got their first Series A raise. I expect these trends to continue and the AI markets to continue to ‘melt up.’

Lots of exciting times ahead for the AI space and 2021 should an interesting year. See you on the other side!

comments powered by Disqus