The Sync Ni team caught up with Nathan Wardlow, Director of Engineering at Highroads to hear all about his current role.
Having graduated in Applied Maths and Physics at Queens, you then went on to complete a PhD specialising in Nuclear Physics in a medical environment. From there you have held a number of data science roles across retail, InsureTech as well as developing AI and Machine learning for media and advertising companies you have come full circle to be working in the Health sector once again. As Director of Engineering at Highroads can you tell us a bit about what this role entails?
Thanks for the introduction. My role at HighRoads pivots a bit from my previous positions which were more focused on data science – HighRoads is first and foremost a software company and my job involves oversight of many of our platform’s features and capabilities.
HighRoads’ main product is a platform for the configuration of US Health Insurance policies and the generation of associated documentation. The company is headquartered in Massachusetts, though the engineering team has a wide geographic spread and is concentrated around our Belfast office.
As part of the engineering leadership team, we look to both improve and add to our existing platform, across the AI/Machine Learning, Cloud-native and more traditional spaces. One of the areas we’ve carved out for focused innovation is Data Analytics and Reporting, which falls under my remit.
In terms of Data Analytics can you share some insights into how the industry has evolved over the last decade and how these have had a positive impact?
One change we probably didn’t see coming is that now (almost) everything has moved off-prem. Cloud service costs and the knowledge/training required for implementation were big barriers to adoption, and the thought of emptying the server room into a skip would have made the CFO cry. That’s turned on its head now, with data and compute power both physically off site. We’ve seen a plethora of “cloud-native” or “cloud-first” organisations spinning up, which would have been pretty hard to imagine 10 years ago. These changes have ushered in the possibility of hybrid/remote working which has an abundance of upsides (particularly noticeable in the covid era).
Storage space has become a non-issue – where previously we may have had to limit the data captured or set strict retention policies, now we often see “capture everything” mindsets with every event or adjustment being tracked. Alongside this has been the rise in unstructured or semi-structured data. Both of these have required adjustments to data analytics practices – tooling has evolved to handle truly BIG data and the potential insights from the mass of available data are staggering.
With these foundational changes to both volume and infrastructure has come the emergence of the Data Engineer and the Data Scientist, where previously we had only DBAs and BI Developers. Data Engineer, Scientist and Analyst job titles have boomed in the past decade, with a wider understanding of the roles, capabilities and anticipated outputs maturing more recently. This has not only provided many people with exciting careers but also given business incredible opportunity for data-driven decision making.
What are the major challenges for AI and Machine learning facing the health insurance sector today?
Red tape is probably the biggest. Insurance is notoriously risk-averse – the sector is only now coming through the digital transformation that other sectors finished 5+ years ago. New and emerging technologies are unknown and scary; convincing stakeholders that the rewards outweigh the potential risks is time consuming and can so often be fruitless.
There’s no shortage of talented people and compute resources in the sector, so from a technological perspective the blocker I’ve seen most often is availability / quality of data. You would think the health/medical/insurance sectors would have an abundance of data and whilst in theory the volume is massive, much of this is still analogue, and what is digitized comes from disparate sources which have many varying ways of collecting and collating it, so leveraging en masse requires incredible investment in cleansing and mapping. Additionally, a majority of data in the space would qualify as PII, so access is restricted by various Data Protection/Privacy legislation. As a society we’re still working through our relationship with data – we’ve put a lot of important safeguards in place recently but are yet to develop the pipelines to efficiently anonymise and then expose public datasets for R&D purposes. It's heart-breaking seeing such a goldmine lie untapped, but I’m hopeful we’ll sort something out here soon.
Finally, there’s the question around adoption – traditionally healthcare decisions are made by rigorously trained professionals, whilst insurance rating tables are derived by actuaries and signed off by underwriters. I’ve seen people from each of these roles rejecting the concept of AI and ML integration from the outset, possibly foreseeing themselves as being “automated out of a job” or made somewhat lesser by the involvement of a model, or not trusting the “black box” as they can’t see the inner workings behind any decision making. The human-machine interface is key: this space in particular benefits from AI and ML input being first to prioritize or pre-select a range of possible options, which are served to the human to facilitate their decision making, keeping expertise and gut feeling part of the process.
If you had a crystal ball what new developments in AI and Data Analytics do you foresee happening in the mid to long term and how do you think these will impact our lives?
As compute power has become more accessible and “paint-by-numbers” modelling became available in recent years, there’s been a flurry of activity where new businesses have tried carving out a niche application of AI or data analysis across many sectors. This has led to a lot of us being exposed to model inferences, good and bad, several times each day.
With the rollout of PII legislation and the exhaustion of these initial first-to-market applications, in the next few years I can see many of these avenues winding down and quality of insight winning out over quantity. The ad-tech space has already iterated through a couple of cycles of population pushback – relevance is key, and many early applications now miss the mark.
If we can get pipelines in place to anonymise and clean data, or provide personal data with appropriate consent, I hope as AI and Data Analytics capabilities mature that our day-to-day activities will become more efficient. We see some of this already; our favourite menu items are called out at the top of the menu in our takeaway apps, or we get a notification “It’s 5pm, do you want to navigate home?” with an embedded map. We make thousands of decisions every day - I don’t want to see decision making taken away entirely, but if the irrelevant options can be pre-screened for me it’ll save me a lot of time and energy.
Modelling human behaviour is an area seeing a lot of R&D investment – the applications here are incredible but something I see a lot of value in is predicting the outcomes of our human decisions and actions; a lot of our anxiety can come from the “what if?” – the more we can understand and predict the effects of our potential actions the more peace of mind we can have at decision time.
Having at various times held positions as a data scientist, VP of product development and currently Director of Engineering what would you say are the most satisfying and rewarding aspects of each of these roles?
As a data scientist I was most excited to see my contributions come to life – there’s nothing quite like seeing a new model perform for the first time (and get it right!), then productionising the model and analysing the improvements when it starts impacting customer journeys.
I was VP of Product Development at a smaller company so had more involvement across the operation – satisfaction often came from taking an idea to the executive team and receiving feedback and sign off, then getting the rest of the development team on board and working alongside them to bring the plan to life.
Currently as Director of Engineering I’m more responsible for a team than my individual contributions. Here I love seeing the team flourishing; individuals doing well and achieving their goals, growing their skillsets, and contributing to the product. Cross-functional relationships are also an important part of my role and it’s incredibly rewarding when I’ve been part of a healthy process from inception to delivery, with both internal and external stakeholders delighted with the finished feature.
Is there anyone you can think of who has been instrumental in the development of your career?
My first line manager, post-academia, Marion Rybnikar, took a chance on someone coming from a non-traditional background into the commercial Analytics space, gave me the space to experiment and try new things, and approached the business with value propositions. She encouraged me to explore and stretch myself, which ultimately led to me seeing my first business data science project (an eCommerce recommendation system) come to fruition. Since then, I’ve continued trying to get as much personal growth as possible out of any position I've held – there’s always room for flexibility beyond the job description; you can often use that grey area to add value to your employer and yourself!
If you could spend a day shadowing one person in the global tech sector who might that be and why?
I’d love to spend some time with Paul Miller of Bethnal Green Ventures – BGV sees all sorts of up-and-coming initiatives seeking acceleration or funding to use technology for positive societal impact. It would be great to be a fly on the wall for a couple of pitches. Paul’s incredibly knowledgeable and is bound to have a wealth of insight into what potential improvements are around the corner, or which worthwhile projects could use some extra development muscle.
For anyone considering a career in AI and Data Analytics what advice would you give to them?
Although public understanding has improved a bit in the past few years, there’s still an aura of mystery surrounding AI – don’t let that put you off, it’s really not that scary! If you’ve any affinity for data and a knack for teasing out solutions to problems, the AI and Data space could be a great fit; at its core you’re trying to teach a computer to do what comes naturally to you (though ideally at a much grander scale), so you can then spend time using the insights to make the best decision.
Working with data can be painful – a good chunk of your time will be spent sifting through, cleaning and mapping source data, but as it’s been likened to “the new gold” or “the new oil”, both need refinement before their true value can be realised. Once adequately prepared you’d be amazed how quickly you can land on actionable results.
If you’re considering it, it’s an easy space to get a feel for – public datasets for tutorial-level problems are readily available so download something that seems interesting and spend a few hours getting your hands dirty – if you like what you find then keep exploring!
