At GlobalLogic, both our consultants and clients love utilising Amazon Web Services (AWS) technologies extensively. This passion has helped us secure a number of highly coveted

accreditations, including Amazon Connect Service Delivery Partner, AWS DevOps Competency Partner and AWS Advanced Consulting Partner.

As part of our partnership, we thoroughly enjoy attending AWS events to keep our fingers on the pulse of the AWS ecosystem. With this in mind, we couldn’t miss out on this years’ AWS re:MARS event in Las Vegas.

MARS stands for:

Machine Learning

Automation

Robotics

Space

This annual conference is one of AWS’ more technical events, with talks from recognised thought leaders and technical experts who are building the future of Artificial Intelligence (AI) and Machine Learning (ML). This conference is the ideal place to apply concepts and technologies that solve customer problems and initiate change within organisations.

There were a number of different formats to the talks, including ‘lecture’ style, ‘whiteboard’ style and ‘workshop style – with increasing amounts of interactivity. You can find a few of the recordings here. No whiteboards sadly.

While this would be a fine opportunity to disappear into the library of MARS content, Sami Alsindi – Lead Data Scientist, Miguel Ribeiro – Senior Data Scientist and Tomasz Walis – have helpfully pulled together their thoughts, discussing favourite speakers, learnings and key highlights.

If you’re ready, let’s jump in.

Sami Alsindi – Lead Data Scientist

re:MARS was held in the very fancy ARIA Resort and Casino hotel – a bottle of water from Room Service costs $27! But, on the other hand, there was a button to automatically draw the curtains, so swings and roundabouts.

Some of the talks were extremely popular, with queues almost exceeding the total seating even 30 minutes before the talks started!

During the event, I found that there were three standout workshops:

1. Amazon Textract
2. Internet of Things (IoT)
3. Taking on a robotic arm

Amazon Textract

Amazon Textract is a managed AWS offering that can extract text from ‘searchable’ PDFs and images as well as perform optical character recognition (OCR) to extract text from non-searchable ones.

I’ve heard about this OCR on many occasions, but this was my first experience working with the different feature sets and testing them out on realistic datasets – it was significantly more powerful than I was expecting!

In less than an hour, we had built a document classification system that automatically detected and sorted a folder of scanned receipts, bank statements and tax documents. It then performed an automatic redaction of Personally Identifiable Information (PII), detecting its presence, and blocking it out on the images themselves. I can’t wait to try it out on some customer use cases!

Internet of Things

During the IoT workshop, I saw massive potential to help utilities’ customers.

In the scenario we were helping to optimise a chemical process to maximise yield. However, the approach used to create the model was a novel one, essentially creating a ‘digital twin’ of the chemical reactor and then simulating different parameter combinations – then working out which is the best.

The most impressive part was the second part – actually deploying it using Amazon SageMaker Neo and Amazon Greengrass onto ‘edge’ devices that run and operate these reactors. Once you have an accurate simulation of the chemical reactors, you can with one click (okay, maybe a few clicks!) automatically deploy these models to all your chemical plants – significantly increasing the yield companywide. Really impressive stuff!

The team takes on a robotic arm

The most fun workshop was focussed on training a robotic arm to, based on simulations run in AWS, pick up and place items in real life. We were the only team to be able to get it working – it was quite challenging and very different from the day job!

The stentrode

As well as workshops, I was really impressed with the keynote speakers. In particular, Synchron’s CEO Thomas Oxley’s talk on how the first recordings made from the human brain in a non-invasive manner allowed a ‘locked-in’ syndrome patient to be able to tweet for the first time.

I have a background in Neuroscience and have always been intrigued in technologies that promise to be able to form impactful Brain Computer Interfaces (BCIs), such as cochlear implants, the Utah Array and, lately, Neuralink.

However, Synchron has achieved something that I have not seen before – the ‘stentrode’. This is a permanent, non-destructive implant (similar to a stent), which records the human brain live in vivo.

A wire containing an electrode is threaded through the brain’s vascular system and then left there permanently – much like a stent in the treatment of atherosclerosis. This didn’t involve any neurosurgery whatsoever, unlike the other options mentioned. The CEO revealed they actually have recordings from a ‘locked-in syndrome’ patient – a rare neurological disorder that paralyses all voluntary muscles, except for those that control the eyes. This recording showed how they were able to communicate via text, which was really inspiring. If you want to see the amazing talk, check out the AWS Events YouTube channel here.

Taking advantage of the work trip, Myself and Miguel stayed an extra night after the conference and explored ‘real’ Vegas. We started away from the Strip in Fremont, which has a really cool vibe and then drove to Death Valley – which was extremely hot in June weather but had some amazing scenery!

All in all, I had a fantastic time at the conference and, if given the opportunity again, would go in a heartbeat.

Ribeiro – Senior Data Scientist

AWS re:MARS was one of the best conferences to get to know new technologies, as well as how they are addressing real world problems. From all the keynotes and presentations, there were two talks to highlight.

Business Transformation through data and ML

The first was PepsiCo’s Business Transformation through data and ML, delivered by Stepan Pushkarev, CTO/CEO at Provectus, and Russell Lamb, Director of Engineering, Data Platform at PepsiCo. This talk was great for our Data Science team, as we’re always looking for new and innovative approaches to put into practice.

PepsiCo started to leverage the data they were holding on-premises and through different data sources – focussing on AI use cases, instead of a data migration and modernisation.

This value-focused approach enabled PepsiCo to start taking value from their ecommerce historical data. With the return on investment (ROI) from this project, PepsiCo then started to invest in other AI opportunities in different business areas – allowing for a larger project which focussed on building their data lake to be used by ML algorithms.

Siemens Energy

The second talk to highlight was Siemens innovative power plant operations using Digital Twins, delivered by Clarissa Raithel, Head of Digital Twin Team at Siemens Energy. Digital Twins was a hot topic in this conference, not just for utilities but also for autonomous driving.

During this talk, Siemens went through their approach to using Digital Twins to monitor their power plants, discussing how Siemens Energy is continuously looking for innovative ways to optimise the power plants they provide to their clients. They are using the AWS Digital Twin Maker solution to monitor all the equipment within a power plant.

This process enables Siemens to gather real-time data on what is happening in the powerplant through a dashboard, and start innovating different areas of the plan:

• Energy management
• Energy consumption
• Systems reliability
• Autonomous operations

With all this data available, Siemens is starting to use AI to:

• Increase their revenues
• Optimise and reduce costs
• Reduce emissions
• Increase brand awareness
• and reduce HSE risks

One project which helped Siemens to reduce costs was the Heat Recovery Steam Generator (HRSG) pipe corrosion, a $200M opportunity. This project enables the team to find patterns in the data that indicates that a failure is about to happen. This then allows the team to be proactive – taking corrective measures and avoiding downtime in production.

This could be a great opportunity for GlobalLogic to explore with our Energy and Utilities customers, as we have the skills and experience in predictive maintenance projects.

Last but not least, ‘JetPack dude’!

This last keynote was delivered by Richard Browning of Gravity Industries, demonstrating the ‘Jet Suit’ he developed, which makes human flight a reality!

This suit is being predominantly used for save and rescue missions by the army, and is also being trialled by emergency services in the UK. Testing is taking place in the Lake District, where someone could be stuck in hard-to-reach areas.

Check out a short video of the suit in action (you might want to cover your ears!)

AWS re:MARS was one of the best tech conferences I’ve been to – a great opportunity to join in-depth workshops, look at the latest developments, and learn how to use these to solve real-world problems.

It was also very insightful to see different approaches that other companies are using to deliver value to their customers.

Tomasz Walis – Senior DevOps Engineer

Conversational AI

Taking an overview of my event experience, keynotes were fun and interesting and

I found the Conversational AI with Amazon Connect talk, delivered by Yoni Barkan, VP Product at Deloitte Digital to be the most engaging.

I work with Amazon Connect and am always looking for new ways to improve customer experience. The talk started by introducing how today’s conversational AI is being implemented and its limitations – where the questions that we ask the AI of today are not being answered correctly and we must go through multiple prompts to get the answers we are looking for.

Deloitte looked at solving this problem by combining two methods:

1. Deterministic
2. Probabilistic

This has resulted in a holistic conversational Natural Language Understanding (NLU) layer.

Deloitte also presented a tool called TrueServe DNA, which has contact centre designer and analytics built in. The talk was very impressive, considering that the contact centre today is heading towards AI, Natural Language Processing (NLP) and automation where customers can experience human like conversation.

Robot dogs

One workshop that deserves special recognition was ‘building the cloud connected robot dogs’ and making them dance.

Before this workshop I had no experience with robots or the Robot Operating System interfaces, however, after only a few minutes it became very natural to use and AWS provided pre-existing environment for attendees to work on. The dance routines were designed in the AWS RoboMaker simulator and then deployed to the robot using AWS IoT Greengrass. It was surprising and interesting to learn that by using AWS Services, the development of robotics becomes more accessible as well as easier.

CodeWhisperer

AWS has also launched new ML Services to help developers write better and faster code. The Amazon CodeWhisperer is a tool that allows developers to generate code using natural language. The tool can examine the code as well as developers' comments and come up with correct code recommendations. Discover more by reading this AWS blog.

Was AWS re:MARS worth attending in 2022? Definitely. There were some amazing talks on how other companies use ML and how they solve problems, and I’m excited to bring my new knowledge back to the GlobalLogic offices.

More than anything, AWS re:MARS offered a unique look into an impressive technology in our daily lives and exciting new tech to come.

Closing remarks

AWS re:MARS was a great and valuable event for GlobalLogic to attend. Alongside the jet pack and robot dogs, this event had some great keynote speakers and workshops for our Data Science team to get stuck into. When looking at what’s to come, it’s clear that IoT technologies for model deployment such as Amazon Greengrass and SageMaker Neo, and and tools that enable development such as CodeWhisperer are going to be popular in building the future of AI and ML.

About the authors

Sami Alsindi – I lead the Data Science consultancy team at GlobalLogic in the UK&I region. I focus mainly on helping clients on their journey towards becoming Data Driven, extracting the untapped value in their datasets using AI/ML. My favourite part of the job is supporting the Data Science team members develop in their careers and seeing them grow more proficient and confident.

Miguel Ribeiro –  I'm Miguel and I work in the Data Science team as a Senior Data Scientist. My favourite part of the job is to understand what's the challenges our clients are facing and create AI/ML solutions to address those challenges. Over my experience, I was able to create AI and ML solutions that enabled my clients to become more data driven and have a more proactive approach.

Tomasz Walis – I work as a Senior DevOps Engineer specialising in Amazon Connect, a call-centre in the Cloud. My favourite part of the job is improving Customer Satisfaction through improving the call flows and making the deployment of call centres as painless as possible.

Behind the scenes

When the conference allowed some downtime, the team had fun exploring the sights and sounds of Vegas. This included a 19 cent win for Sami (or $9.81 loss from a $10.00 investment depending how you look at it)…

…and a round of roulette for Miguel. Sadly, betting everything on Cristiano Ronaldo’s shirt number didn’t work out too well.

In conclusion, we’d say the highlight of the trip was when [REDACTED] did [REDACTED].

Confused?

Probably for the best. What happens in Vegas stays in Vegas after all!