High Impact AI healthcare jobs from my experience

Dr Richard Freeman
9 min readJun 13, 2022


Credits Richard Freeman, PhD

With a PhD in machine learning and NLP, 18 years experience including 3 as CTO for Data in healthcare, I wanted to share my story as if we were having a conversation over drinks — a candid discussion about how your tech skills can make a huge difference in healthcare and life sciences. This is based on my personal experience which includes discussions with hundreds of healthcare experts, consultants, developers, doctors, recruiters, clients, and data scientists.

I’ve always been a big advocate of tech for good, having spent six years at the Crowdfunding for good platform JustGiving, where I deployed low latency data science models at a scale of 26M users. In 2013 this was very hard to do. See my blog on how you can get involved in tech for good. I could have worked anywhere given my skills in AWS, solutions architecture, big data, machine learning, natural language processing, deep learning, graph analytics, Serverless, and streaming analytics, but I needed an exciting tech challenge with an innovative company. What I wanted to do next was prioritise healthcare over other industry sectors.

Coming from outside of healthcare and life sciences, what I found once ‘on the ground’ was very different from what I expected. In my view, most of us will go through a similar thought process and jump straight into a healthcare organisation without doing much background research. If you really want to be in the sector and are good at data science (or even if you think that you are the best), you will need much domain knowledge. Make sure you are ready to invest the time and are passionate about healthcare, life sciences and the human body.

Let’s first exclude traditional pharmaceutical and medtech IT functions that would be the same as other industries that do not work on patient data. This includes AI-powered chatbots, advertising, and marketing. After my time at JustGiving that was later acquired by Nasdaq listed BlackBaud, backed in 2019 I received several job offers very quickly, one was for a wearable app that tracked weight loss with a smart watch, another focused on finding patients for clinical trials, but I felt these were just evolving existing tech without breaking new ground. And others were too narrow, focusing on using data science to diagnose things like bone fractures, which is simply a computer vision problem. While interesting, in my view this is something most data scientists could solve using a pre-trained deep learning model or even existing libraries in four lines of code. So, after a few weeks of work what will you have to do? Perhaps support the sales team, but they are unlikely to have much activity either given they will be selling only a single diagnosis solution that is easy for competitors to copy.

Let’s now think about the patient side. I think there are many developers that have big ideas about forming healthcare startups to revolutionize ‘broken healthcare delivery systems’ or ‘help millions of patients’. However, they quickly realise that providing either medical advice, diagnostics, or even consultation to a patient is a highly regulated field and cannot simply be hacked with a data science model from Kaggle. So what they tend to do is switch to the next best thing where patients become consumers, which is a wellness app or life/habit/food/mindfulness coaching apps — or stay on the borderline like a genetic startup I spoke with, that provided health info accompanied by a big disclaimer that it should not be considered a medical diagnosis (and may be for entertainment purposes only). Perhaps this is a generalisation, but do a search for ‘wellness’ or ‘coach’ on your app store to see for yourself. This may seem controversial, but in my view these apps do very little for humanity in terms of solving fundamental problems, there are of course some exceptions like BlueHeart and other psychological and medical apps, but this space is now very crowded and you missed this wave already, sorry. Please don’t get me started on the number of diagnostic chatbots that are scripted or powered by NLP training sets which have a very low barrier to entry now that AWS has several managed services (and has for years) like chatbots with conversational AI called LEX or the more recent chatbot which offers easy monitoring, operating, and troubleshooting of AWS workloads in the chat channel.

Other areas you will come across include finding patients for clinical trials, which is critical to most pharmaceuticals, especially when these concern rare diseases. However, this is a challenge that has already had a lot of data science investment and to some degree is a solved problem that has become more a data sourcing challenge.

In my view there are three areas you should work in healthcare if you want to apply AI for good and maximise the benefits for humanity.

1) Robotic surgery.

— use computer vision and robotics for fully automating surgery

Although common in science fiction movies like the Fifth Element, Ender’s Game, and Prometheus, the realization of a MedPod or robot surgeon performing a full range of diagnostics and therapeutics is still decades away. Think about the precision of robots used in mass production manufacturing or 3D printing today and the potential to transfer that more broadly into healthcare.

I was recently challenged to talk about the benefits of AI, as there seems to be a big fear around AI taking over physician jobs. But imagine that in 2035 your loved one had a serious accident and you had to choose between a 99.999% surgery success rate based on one million procedures, or an expert human surgeon with a 95% success rate based on 500 procedures, a much longer recovery time, and more expense. Which one would you select? Imagine the human doctor who is certifying the robot surgeon will tell the patients or their loved ones that it is a ‘fully automated, FDA approved, and repeatable process that has been tested on millions of patients over a decade and has seen trillions of different simulations of almost anything that can happen and will react in nanoseconds’.

Thinking further on nanotechnology, imagine simply swallowing a pill or injecting a tiny robot into your bloodstream that will explore your body and remove any tumors or cancer cells without making any incisions at all, or maybe one that will massage your muscles or activate grey matter (think brain hacking). Likewise in brain–machine interfaces (BMIs), even Elon musk is on this topic with Neuralink which wants to automate the implanting. “We ultimately want this robot to do essentially the entire surgery — so everything from incision, removing the skull, inserting electrodes, placing the device and then closing things up,” said Musk during a live event. “We want to have a fully automated system.”

Why is it challenging?

This is the holy grail. It is very complicated robotics and involves complex real-time computer vision and sensors. I would say it is even more complex than self-driving cars due to the many possible scenarios that can happen with the human body. Will the computer vision, sensors, and robotics involved be able to mimic a human surgeon? Also think about the liability and ethics side if things go wrong.

Who to work for?

This is still a very active area but in my view there are limited jobs available and you will need a background in mechanical engineering, mechatronics, or computer vision skills to be considered. In addition, you need to look at this in isolation until the technology advances to a point that it is fully merged into an autonomous surgeon.

  • Computer vision diagnosis
  • Robotic assisted surgery like the Da Vinci Surgery, or CMR Surgical
  • Medtech and implantables still very much in the research domain
  • Nanotechnology still very much in the research domain

2) Drug discovery

— using data science to find new drugs

When I was looking for a job in the healthcare field, drug discovery sounded the most interesting. However, after speaking with recruiters and companies I realised that to be involved in the most drug discovery areas you need to have a solid background in bioinformatics, chemistry, and biology. Essentially, a lot of the work is simulation based, and you would run simulations on large scale High Performance Computing or Graphics Processing Unit servers.

Why is it complex?

  • You need deep domain knowledge overlapped by strong data science skills
  • You will probably need a PhD in the domain, along with a few postdocs and many journal publications in the area. And then be backed by solid experience in industry.

What if I want to work in this area?

If you discount the companies that help drug discovery, some touch these such as using NLP to analyse open medical publications, establish biomarkers, and the ones that are actually doing drug design examples include Healx, BenevolentAI, Exscientia and more that do.

3) Global Healthcare marketplace

— connecting healthcare suppliers with hospitals by products/services and outcomes

Healthcare procurement is a forgotten area of AI-application, yet it is very complex and exciting with high impact potential. Much of the focus concerning use of AI has been on patient diagnostics and monitoring, drug discovery, or business operations, but what about all of the actual products needed to enable care delivery? Without this all the R&D outputs never gets deployed and goes to waste. Does the hospital or buyer really know all the suppliers in the market? The answer is no, not only in Europe but most of the world.

During the pandemic, delays in procurement have meant that people have died unnecessarily. During the first COVID-19 wave there was a global shortage of face masks and PPE due to an over reliance on the just-in-time model, which led to preventable death and many heart breaking scenes. For those in India during April 2021, there was also a shortage of oxygen cylinders leading to even more dramatic and despairing scenes, which again is a procurement issue.

Why it is challenging?

  • There is a huge amount of unstructured data in multiple languages, with variety, velocity and scaling issues.
  • There is a large amount of data exchanges and integration needed on both the supplier and buyer side

What is the AI angle?

All of the AI techniques can be utilised: Patient prediction using multivariable time series, supplier product matching with buyers using graph analytics, supplier competitor analytics, auto-catalogue classification of healthcare products using NLP, price predictions using game theory and machine learning, content translations using neural machine translation, medical named entity extraction using NLP, document and content processing using big data, and many more.

Lets further explore the supply chain network, as performed in a research-funded project with Vamstar and expert NLP PhDs from the University of Sheffield. We essentially used NLP and machine translation to build up a global network that connects all countries, supplier products/services, and buyer like hospitals. It is broken down by product features and merged with medical ontologies, clinical trial data, opportunity data, and geolocation data making it a huge network. We then use complex graph theory algorithms to match suppliers with buyers that have the most suitable products, find the best opportunities for suppliers, infer new relationships based on diseases, and identify the strength of these relationships. Adding time series forecasting as the network evolves allows us to have a progressive view on products and supply. Finally, if you add machine learning for classifications and game theory in pricing simulations you have suddenly covered the major areas of data science. Then if you think about a truly self-sustaining marketplace as something similar to what automated algorithmic trading did to investment banking, you can imagine real-time and integrated systems that respond to supply and demand elastically and with great efficiency. Through automation, humans previously involved in the area can now focus on more high-value, rewarding tasks such as querying the network graph, running data science experiments, and patient care.

If you want to work on the AI-powered healthcare marketplace:

Vamstar (disclaimer I’m a co-founder) are completely healthcare focused as this is a technically challenging space in data science that need a specialisation.


Good luck in your potential career in healthcare/life sciences and the benefits you will bring to humanity with the positive side of AI. You now know the three areas of high impact that are robotic surgery, drug discovery and healthcare marketplace, you can use as a springboard to find out more if you make the jump. Feel free to post comments or suggestions, this represents my personal views and journey and is not a Vamstar company position.



Dr Richard Freeman

Author, Advisor, Co-founder & CTO Data @ Vamstar, Series-A funded startup