Solliance makes headlines with cryptocurrency news analysis platform powered by Azure Machine Learning, PyTorch
This blog post was originally posted on: https://customers.microsoft.com/en-us/story/1457351257191063636-solliance-partner-professional-services-pytorch-on-azure
Solliance delivers cutting-edge solutions that fill gaps across a wide variety of industries. Through its recent collaboration with Baseline, Solliance revolutionizes the cryptocurrency trading experience, extracting news insights from more than 150,000 global sources in near real time. To manage Baseline workloads, Solliance brought Microsoft Azure Machine Learning and PyTorch together for maximum processing power and deep learning capabilities. The result: investors can get under the headlines and see which specific news metrics are moving the volatile crypto market to make more informed trading decisions, while Baseline can release new features in weeks instead of months.
“With Azure AI and PyTorch, we combined focused applications of AI with journalistic processes and financial intelligence, yielding a solution that is unique in the market and valuable for cryptocurrency investors.”
— Zoiner Tejada: CEO, Solliance and CTO, Baseline
A game-changing solution
Depending on who you ask, cryptocurrency markets are complex, controversial, scary, exciting, good bets, the Wild West, serious business, a flash in the pan, and the wave of the future. Everybody’s right to an opinion notwithstanding, cryptocurrency sits famously at the current leading edge of financial innovation, and one thing is clear. Worldwide, cryptocurrency is worth almost a trillion dollars, and the big players have noticed.
Solliance, an organization that’s no stranger to innovation, develops and delivers expert consultancy and cloud, data, and AI solutions that make it easier for business professionals to make an impact. Often, the company builds new solutions for transformative startups across sectors from media to finance. One such startup is Baseline, a crypto asset management and analytics platform that institutional investors can use to capture unstructured data in cryptocurrency news stories and process it into quantified data they can follow and trade against.
With Baseline, investors aggregate information from more than 150,000 global sources, including traditional and trade publications and social media, in near real time. They can then combine news indicators with crypto asset analytics, values, and metrics, and compare reports from multiple outlets to get accurate insights and ultimately quantify the relationship between news events and crypto market movement. The platform aggregates all of this information into “microfacts,” scores them for credibility, and tracks them over time as new information emerges.
The hard part
News sources are disparate, dynamic, often siloed, and expanding, and the content is hard to quantify. The most difficult part is separating the news that moves markets from the noise — in an environment where offhand tweets from celebrity billionaires can send markets soaring through the stratosphere.
Baseline needed an AI platform that would make it easy to get started fast, iterate quickly, try things, and break down technical obstacles. It built one with Microsoft Azure Machine Learning and PyTorch on Azure for the processing power, scalability, and efficiency it needed to meet its ambitious technology and business goals.
“We started down this path to see what we could do with critical news information in an extremely volatile and high-stakes market,” says George Howell, CEO of Baseline. “With Azure AI and PyTorch, we feel like we found some pretty innovative ways to help bring clarity to a very crowded information ecosystem.”
End-to-end data science
Baseline wanted a set of machine learning services that would support end-to-end data science processes centered around natural language processing, large amounts of roll computing capacity available to scale up or down as needed, a development ecosystem centered around deep learning models, and powerful natural language processing models to use as building blocks for more advanced Baseline features.
To extract news content at the sentence level and evaluate, monitor, and track how it evolves over time, Baseline uses Azure Machine Learning to build, train, and run models at cloud scale to handle immense data volumes. By combining Azure Machine Learning and neural networks based in PyTorch, Baseline can process multiple news events simultaneously.
“A big part of our natural language processing is based on transformer models, so PyTorch is hugely valuable to us as the leading edge for those models,” says Zoiner Tejada, CEO at Solliance and CTO at Baseline. “And with Azure Machine Learning, we can improve the experience with PyTorch-based natural language processing models and make each solution more valuable for us.”
With deep learning based on PyTorch and powerful, pretrained machine learning models in Azure Machine Learning, Baseline has addressed the technical obstacles it faced in a way that’s efficient and easy for both now and the future. “When we need to calculate similarities between very large sets of vectors resulting from hundreds of thousands of sentences, the compute capabilities and all the support that we get from managing these workloads in Azure Machine Learning really shines,” says Ciprian Jichici, Chief Data Scientist at Solliance and VP of AI at Baseline.
Easier, faster, quantifiable news analysis
Since starting this journey, Baseline has accelerated time to value and sees limitless potential for what it might achieve in enhancing the crypto trading space. “A lot of the architectural thinking around Baseline was possible because we knew we had this power available to us,” says Jichici. “I would have never dared to think about some of the approaches that we have today without something like Azure Machine Learning underneath us, or without the very powerful PyTorch-based natural language processing models that we knew we could use as building blocks for the more complex types of modeling that we do in Baseline.”
Jichici recalls multiple instances of the Baseline team brainstorming complex new features and producing experimental implementations in a matter of weeks — a speed that is extremely valuable in an industry where things happen very quickly. “We don’t have the luxury of spending six months to bring something new to the table,” he notes. “In that time, that new thing could already be obsolete in the space.”
Not only do investors now save time by not having to consume so much media on their own, but Baseline also makes that information concrete and can indicate whether there’s a strong consensus score among industry newsmakers’ varying predictions. The standard timeline for an individual investor analyzing small news events is anywhere from three to eight hours, but Baseline can complete its processing in under five minutes. “These are the early days, so we want to get that processing time down to seconds eventually,” says Tejada.
Given its advanced features, capabilities, and intelligence, there is no doubt Baseline will become a critical tool in crypto currency investors’ trading strategies moving forward, ensuring they are always prepared for the next breaking story.
“With Azure AI and PyTorch, we combined focused applications of AI with journalistic processes and financial intelligence, yielding a solution that is unique in the market and valuable for cryptocurrency investors,” says Tejada.
“We started down this path to see what we could do with critical news information in an extremely volatile and high-stakes market. With Azure AI and PyTorch, we feel like we found some pretty innovative ways to help bring clarity to a very crowded information ecosystem.”
— George Howell: CEO, Baseline
“When we need to calculate similarities between very large sets of vectors resulting from hundreds of thousands of sentences, the compute capabilities and all the support that we get from managing these workloads in Azure Machine Learning really shines.”
— Ciprian Jichici: Chief Data Scientist, Solliance and VP of AI, Baseline