On December 7, 2022, we got the chance to bring the data science community in NYC back together with the Data Science Salon NYC. The last in-person event we had hosted in the Big Apple was in 2019 and we had missed seeing everybody face-to-face so much!
The audience at DSS NYC 2022
The event was a huge success, with over 150 attendees joining us live at The Theater Center on Broadway, and many more virtually.
Data Science Salon NYC featured 20 speakers from leading companies in finance and technology who shared their experience and best practices in applying AI and machine learning to the industry. Read here about our main takeaways from the event!
Data Science Salon NYC started with the presentation of Sajjad Farahani - Lead Data Scientist at S&P Global Market Intelligence. He talked about the application of machine learning for spreading financial statements.
Sajjad Farahani at DSS NYC 2022
Takeaway #1: In order to solve a particular problem with machine learning, a variety of techniques need to be assessed thoughtfully with the target context in mind. There is no prescriptive method that is purely tied to a particular class of algorithms; the risk context always needs to be kept in mind in order to assess the tradeoffs.
The next speaker was Hongfei Li – Head of Analytics at Point72. In her talk, Hongfei introduced alternative data, its applications in finance, interesting use cases, challenges, and more.
Takeaway #2: An amazing fact for data scientists: the number of alternative data providers is more than 20 times larger now than it was 30 years ago!
Justin Fine, Director of Field Engineering at Katana Graph talked about fraudulent actor analysis methods enriched by graph analytics and AI features.
Audience questions at DSS NYC 2022
Takeaway #3: Tools like Katana Graph can be used to prevent consumer fraud in real-time by detecting suspicious transaction patterns as fast and as accurately as possible.
Chong Dang, Lead Machine Learning Engineer at Freddie Mac walked the audience through an NLP-based end2end pipeline with Amazon Textract and Comprehend to analyze and extract the key-value pairs in scanned pdf documents.
Takeaway #4: Document AI Pipeline (DAP), an AI-powered mortgage solution that uses Amazon Textract and Comprehend (NER) can be used to perform document recognition, data extraction, and the analysis of income, assets, and property appraisal in order to increase lender productivity, drive operational efficiency, and reduce costs, resulting in higher customer satisfaction.
The next talk was presented by Kaili Li, Sr Manager, Data Science at T-mobile, and Michael Kortering, Chief Credit Officer at Reprise Financial. In their co-presentation, they covered the strategic thinking process of how to leverage machine learning to gain market advantage in the credit underwriting space.
Kaili Li and Michael Kortering at DSS NYC 2022
Takeaway #5: Always find a technique for the problem, not a problem for the technique.
Jayeeta Putatunda, Senior Data Scientist at Fitch Ratings, gave an overview of AI in finance – where it is and the path forward. She also spoke about preparing for the NLP revolution and extracting value as a data-driven organization.
Takeaway #6: Always review the AI use case and dataset with a business sense.
A great lesson about making machine learning pipelines scalable was given by Vincent David, Senior Director of Machine Learning at Capital One. He covered best practices for integrating ML capabilities into an enterprise-grade experience and showed that success with MLOps is achievable.
1. Focus on SDK standard, a unified stack, and incentivize reuse
2. Prioritize a clear set of set of use cases and stakeholders and build for those folks first
3. Focus on monitoring and logging. Take requirements and make solutions you can take pride in.
Erin Stanton at DSS NYC 2022
In her talk, Erin Stanton, Global Head of Portfolio and Trading Analytic Client Support at Virtu Financial, walked us through what has worked, and more importantly, what has not, within the trading analytics space.
Takeaway #8: When it comes to building trust through transparency, while also gaining the required skills to train machine learning models, it is best to start with a small, well-defined goal. Also, loop in stakeholders early and consistently.
Sr Researcher at the Federal Reserve Bank of NY, Harry Mendell, discussed how we can use AI and ML at scale to solve real-world problems in financial regulatory compliance.
Takeaway #9: Machine learning is very effective in detecting bad actors, so fraud detection is essentially solved by looking at a very large number of variables provided for each transaction, and supervised deep learning works quite well to detect market manipulation, insider trading and other trading violations.
Networking at DSS NYC 2022