Deep Learning in Finance Summit schedule

The amount of textual data on the Web is enormous and it is growing at a rapid pace. However, in comparison to structured in-house data, this is a data source that is hardly utilized for knowledge extraction at all. While there is no shortage of unstructured data, the challenge comes in accessing that data and transforming it into something usable and actionable. In this talk, we will describe how Mito.ai use machine learning and open linked data to tame the continuous stream of unstructured web data. The final product is a personal stock trading assistant where data driven insight about companies, products, employees and markets is made available in a chat based interface.

Jon Espen Ingvaldsen is a passionate computer scientist and skilled software engineer. Throughout his career has had one foot in academia and the other in the industry. Ingvaldsen has a Ph.D. degree related to analysis and monitoring of large data streams. After his PhD, he has worked as a consultant and software engineer for several industrial projects and Postdoc research fellow at NTNU.

As CTO and Co-Founder of Mito.ai, he is leading the development of the next generation trading platform where media monitoring and trading is combined in one chat-based and mobile interface.

Interpreting Machine Learning Models

Big data and machine learning are becoming central parts of the business for many organisations in the public and private sectors. This is driven by the continuous growth and availability of data that empowers the organisations to make better data driven decisions. To support the business and the users of the machine learning models, it is key to explain why the algorithm made a certain decision. This way machine and people can work together to tackle important issues like Fraud. In addition, with the imminent coming of the new regulation on data protection, today model interpretability is becoming more important than ever. Many machine learning algorithms like gradient boosted trees and deep learning have been traditionally called ‘black-box’ algorithms, because it seems unclear the process they undertake to make a certain decision. In particular, deep learning involves feeding information through non-linear neural networks that classify data based on the outputs from previous layers, making it very difficult to understand the reasons for the decisions made. In this talk, I will discuss different implementations that allow us to understand why an algorithm makes a certain decision at an observation level. I will show how we can use these tools in public datasets and then describe how we use them in insurance.

Soledad is a Lead Data Scientist at LV=, with 2+ years of experience in data science and analytics in the financial sector, and 10+ years of experience in scientific research in academia. She is passionate about extracting meaningful information from data and supporting institutions make solid and reliable data driven decisions. At LV=, Soledad and the data science team are leading the implementation of machine learning across the multiple company business areas. Having transitioned from academia to data science, Soledad is passionate about enabling and facilitating data scientists and academics transition into the field, and helping data scientists increase their breath of knowledge. During the last year, Soledad shared insight in blogs and talks in the data science community. She also created 2 online courses on machine learning now live in Udemy, which have enrolled 350+ students from several parts of the world in just under 3 months.

Recent Developments in Deep Learning in Finance

This talk aims to provide a literature survey of published use cases and research papers on use of machine learning in finance and how it is helping re-focus the financial sector to its fundamental purpose. The discussion will in particular focus on recent developments in deep learning applications and put a spotlight on some of the relevant research in deep learning and reinforcement learning. Other aspects like generating synthetic data, text analytics, transfer learning and explainability of deep learning models will also be discussed. This talk will conclude with update on evolving regulatory landscape and some ethical questions about use of these models.

Harsh currently works with Morgan Stanley in Quant Analytics Group. He started his career as a programmer focussed on developing data driven algos in the areas of speech recognition, image processing and bioinformatics. He then moved to financial risk management and over the last 12 years has worked in various roles through the life cycle of models. In these roles, he has been continuously enthusiastic to applying machine learning in problems related to behavioural assumptions, data quality, recommender systems, model benchmarking and text analytics. His current role requires him reviewing all Machine Learning models used by the firm and providing direction to shaping AIML governance framework and strategy. He is also a visiting lecturer with universities and training institutions.

A look into leading-edge statistical methods to tackle real-world problems, which today means applying machine learning and neural networks to large-scale, multi-structured data sets.

Michael is the Chief Science Officer of Prudential plc. He joined Prudential last year from Silicon Valley based Pivotal Labs where he led the Data Science team. His experience lies in the application of artificial intelligence methods to large-scale, multi-structured data sets, in particular neural network based deep learning techniques. Michael previously founded and sold a ‘Silicon Roundabout’ based startup and prior to that was a partner at a major consulting firm. Michael holds a PhD in theoretical physics from the University of Cambridge and is a Fellow of the Royal Statistical Society.

Practical Aspects of Applying Deep Learning for Market Making

Deep Learning has been shown to outperform traditional methods in many learning tasks such as image and voice recognition, but its role in processing financial datasets is yet to be fully discovered. In this talk, I will share practical insights about applying Deep Learning for different aspects of market making. I will discuss some of the unique challenges and tradeoffs of this field.

I am a senior data scientist at Citi, working at the Technology Innovation Center in Tel Aviv. I apply Deep Learning to financial datasets to build models for different business units in Citi. Before joining Citi I worked as a senior quantitative researcher for a large hedge fund and before that as an algorithm engineer in the field of biomedical signal processing. I hold a PhD in Biomedical Engineering.

The Lego-Ization of Deep Learning

We all know that deep learning can be used to solve a lot of real world problems, but can it be effectively used by domain experts, without a data science background? This presentation covers the use of DL to create a risk scoring system and show how effective it was without the use of a dedicated data science team.

Rich Casselberry is a Risk Technologist at Liberty Mutual. In this role he helps determine risk of current and emerging technology, such as Internet of Things, Chatbots, and machine learning. Prior to Liberty Rich managed the IT operations for Extreme Networks, Enterasys Networks and Cabletron Systems. In this role, he managed the day to day IT infrastructure and security as well as helping to guide the industry by working with the research and development teams to create new solutions to solve real IT needs.

Signal Processing on Social Media: Theory and Evidence from Financial Markets

We analyze the processing of information from social media and news media, using a unique dataset on financial markets. We find patterns consistent with a theory of social media as an “echo chamber”: Social networks repeat information, but boundedly rational investors interpret repeated signals as new information. This is based on the empirical finding that stocks with high social media coverage experience high subsequent volatility and trading activity, while high news media coverage predicts low volatility and trading activity. Alternative mechanisms based on private information, investor disagreement, uncertainty shocks, and other behavioral biases are not consistent with the data.

Peiran Jiao is a research fellow at Nuffield College and the Department of Economics, University of Oxford. He holds a PhD in Economics from Claremont Graduate University, CA, USA. His main research interests are behavioural and experimental economics and finance. His current projects focus on (1) experience-based learning in financial decision-making and game theoretic interactions, (2) information processing from the news and social media with implications in asset-pricing, macroeconomic outcomes and politics; (3) individual investors’ ability to solve complicated problems in financial markets. He also worked on neurofinance, studying the effects of hormone on trading behaviour and market dynamics.

LSTMs and Multidimensional Time Series Data for Enhancing Investment Insight

Long-Short Term Memory (LSTM) neural networks are a relatively unexplored area in the field of finance. Alternative datasets for driving investment insights are also something that is only just beginning to get traction amongst asset managers. Using LSTMs on multidimensional time series and sequential finance data as well as non-finance alternative data such as Internet of Things (IoT) sensors we can attempt to derive hidden values from such datasets to further enhance investment strategies and gain an information advantage over the market.

I am Head of Quantitative Research Development in the Global Investment Strategy team at HSBC Global Asset Management in London, where I specialise in building technology products for our strategic investment vision. I’m also an independent Artificial Intelligence/Software Development Consultant and Entrepreneur in using alternative data for investment insight, I hold a deep interest in using AI in the FinTech space.

On the weekends I am a professional skydiver, skydiving coach and current holder of British skydiving records.

The Importance of Emotion Recognition in Business Relations

When interacting with other people, we tend to give indications to help them understand how we are feeling. These indicators involve emotional expressions either through body language such as several facial expressions, or through speech emotion recognition such as various vocal tones. In particular, with the recent advancements in artificial intelligence, it is now possible to implement an emotion recognizer that helps classify the different emotions people are showing in real time. Whether through facial expressions detection or speech emotion recognition, introducing smart emotion recognizers in business relations can help large businesses better understand their clients and henceforth improve the services they have to offer.

Charbel is a Computer Engineer (B.Eng) with a Master’s degree in Artificial Intelligence from the University of Paris VI (UPMC – Sorbonne Universités). With 2 years of experience, Charbel has worked on different projects in Machine Learning and Artificial Intelligence including an emotion recognizer through speech and face detection, a prototype classifier for a collaborative e-meeting platform, a multilingual speech recognizer & machine translator for smartwatches, and a word-sense disambiguation model. Residing in Paris, he is currently working as an AI specialist and data scientist at the innovation pole of Crédit Agricole, one of the largest bank institutions in France.

Deep Portfolio Theory

By applying deep learning (DL) to classic portfolio optimisation, we show how to capture (or ‘learn’) non-linear features which are invisible to classic portfolio theory. We develop a self-contained four step routine of encode, calibrate, validate, and verify to improve ex-post performance in index tracking and benchmark outperformance.

Jan Hendrik Witte is a numerical analyst who has developed a number of new numerical algorithms in the area of optimal stochastic control. Since leaving academia, Witte has been working as an FX quant in finance. Witte is generally interested in the areas of numerical mathematical finance, systematic trading, and portfolio optimisation. Together with GreyMaths, Witte is building deep learning technologies for the use in trading and investing.

Five years ago Adam talked here at ReWork on the future of AI in banking, he laid out a roadmap of things to come - previewing China’s Smile to Pay (which shortly after ballooned to 400 million users), smart contracts, personalised banking and other initiatives that have come to fruition. That said, Adam had also warned on the grave consequences of failure and significant challenges to arise both practical and regulatory which closely describes the state of quagmire we now find ourselves in.

From Alexa telling a 10-year old girl to touch a live plug with a penny, to regulatory breaches such as Uber test driving autonomous cars without state permission, running 6 red lights as a result.

Finance hasn’t been spared the controversy either, with multiple institutions losing significant wealth from poorly deployed AI, to colossal investment wastage by backing the wrong AI sectors.

In this talk Adam will share key insights & learnings on how organisations can better traverse these pitfalls by designing, building and deploying finance driven AI that is both sustainable and cost effective. He will also layout key milestones to enable future proofing imminent threats of supply chain failures and global talent shortages so that we can navigate the next five years of AI in finance with confidence.

Adam McMurchie is the Lead Cloud Data Engineer at NatWest. He was previously the leader in Devops and an A.I expert working in the banks SAO platform on the forefront of technology development in finance. With a broad exposure to a range of technologies, Adam drives an ethos of simplification, cloud agnosticism and specialises in spotting the next trends in fin tech. Additionally, Adam also has a background in science with a physics degree specialising in NeuroComputing and is a polyglot linguist & seasoned translator. Adam has pooled these skills to deliver full stack novel solutions from tensor flow driven mobile apps, to personalized banking chatbots. Adam also develops apps designed around the ethos of Social Utility, including Flood/Storm reporting, EV Vehicle bay monitoring and preservation of endangered languages.

Five years ago Adam talked here at ReWork on the future of AI in banking, he laid out a roadmap of things to come - previewing China’s Smile to Pay (which shortly after ballooned to 400 million users), smart contracts, personalised banking and other initiatives that have come to fruition. That said, Adam had also warned on the grave consequences of failure and significant challenges to arise both practical and regulatory which closely describes the state of quagmire we now find ourselves in.

From Alexa telling a 10-year old girl to touch a live plug with a penny, to regulatory breaches such as Uber test driving autonomous cars without state permission, running 6 red lights as a result.

Finance hasn’t been spared the controversy either, with multiple institutions losing significant wealth from poorly deployed AI, to colossal investment wastage by backing the wrong AI sectors.

In this talk Adam will share key insights & learnings on how organisations can better traverse these pitfalls by designing, building and deploying finance driven AI that is both sustainable and cost effective. He will also layout key milestones to enable future proofing imminent threats of supply chain failures and global talent shortages so that we can navigate the next five years of AI in finance with confidence.

Adam McMurchie is the Lead Cloud Data Engineer at NatWest. He was previously the leader in Devops and an A.I expert working in the banks SAO platform on the forefront of technology development in finance. With a broad exposure to a range of technologies, Adam drives an ethos of simplification, cloud agnosticism and specialises in spotting the next trends in fin tech. Additionally, Adam also has a background in science with a physics degree specialising in NeuroComputing and is a polyglot linguist & seasoned translator. Adam has pooled these skills to deliver full stack novel solutions from tensor flow driven mobile apps, to personalized banking chatbots. Adam also develops apps designed around the ethos of Social Utility, including Flood/Storm reporting, EV Vehicle bay monitoring and preservation of endangered languages.

Dr Saied Abedi is the founder of Mathficast Software Services UK Limited, the big data predictive analytics company whose Machine Learning-powered stock and commodity market prediction software applications are currently Google page-one hits globally (Please search for "Software to Predict Stock Market" and watch the Mathficast YouTube video). Saied is a prolific Iranian-British inventor holding more than 22 granted US and UK patents in intelligent algorithms. Since September 2015, Saied has been the product owner and manager of Tesco's £100m grocery recommendations program, working closely with dunnhumby. He led Tesco/dunnhumby multi-discipline substitutes program and as a result of his algorithmic initiatives, Tesco substitutes product has been named as the best in the UK market during the fiscal year 2016/2017 (according to Customer Insight reports) while under-spending the Refund-The-Difference budget and achieving the lowest customer rejection ratio of 6.9% in Tesco's history and the highest customer satisfaction. Saied currently leads his elite Tesco Recommendations Service (code-named T-Rex) team based in Farringdon, City of London, responsible for development of Tesco's disruptive graph-based T-Rex platform, currently achieving 18% basket-add conversion rate. Prior to joining Tesco as the Head of Algorithms at import.io, Saied has led and contributed to the development of Magic, the world's first fully automated web data extraction platform. Before joining import.io, Saied has served as a lead innovator and Principal Researcher at Fujitsu Laboratories of Europe Limited, UK for 11 years, running number of multi-million R&D projects successfully from initial ideation to full products or Proof of Concept. Saied has received his BSc in Telecommunications from the Sharif University of Technology, Tehran, Iran, MSc in Machine Intelligence and Computer Vision and a PhD in Mobile Communications-NP-Hard problems both from the University of Surrey, UK. Dr Abedi's biography has featured in many Who's Who publications and the prestigious Cambridge Blue Book (2005 edition) and the Cambridge International Biography Directory. Saied is a keen table football player (and a former regional table football champion). To see more details please visit https://uk.linkedin.com/in/saied-abedi-a179678.

Ajit is a Director in PwC's Fintech and Digital Banking practice where he focuses on safe and effective commercialisation of AI and Blockchain in banking. Prior to PwC, Ajit worked as a Director at Barclays and Goldman Sachs in investment banking risk and compliance technology where he led large global teams to drive strategic regulatory and technology change programs. Apart from banking risk and regulation, Ajit also has extensive hands on expertise in advanced analytics, network engineering, distributed computing and quantitative modelling.