Few seats on the Machine Learning table span each technical skills, problem solving and business acumen like Directors of Machine Learning
Directors of Machine Learning and/or Data Science are sometimes expected to design ML systems, have deep knowledge of mathematics, familiarity with ML frameworks, wealthy data architecture understanding, experience applying ML to real-world applications, solid communication skills, and infrequently expected to maintain on top of industry developments. A tall order!
For these reasons, we’ve tapped into this unique group of ML Directors for a series of articles highlighting their thoughts on current ML insights and industry trends starting from Healthcare to Finance, eCommerce, SaaS, Research, Media, and more. For instance, one Director will note how ML might be used to scale back empty deadheading truck driving (which occurs ~20% of the time) right down to just 19% would cut carbon emissions by ~100,000 Americans. Note: That is back of napkin math, done by an ex-rocket Scientist nevertheless, so we’ll take it.
In this primary installment, you’ll hear from a researcher (who’s using ground penetrating radar to detect buried landmines), an ex-Rocket Scientist, a Dzongkha fluent amateur gamer (Kuzu = Hello!), an ex-van living Scientist, a high-performance Data Science team coach who’s still very hands-on, an information practitioner who values relationships, family, dogs, and pizza. —All of whom are currently Directors of Machine Learning with wealthy field insights.
🚀 Let’s meet some top Machine Learning Directors and listen to what they need to say about Machine Learning’s impact on their prospective industries:
Archi Mitra
Background: Bringing balance to the promise of ML for business. People over Process. Strategy over Hope. AI Ethics over AI Profits. Brown Latest Yorker.
Fun Fact: I can speak Dzongkha (google it!) and am a supporter of Youth for Seva.
Buzzfeed: An American Web media, news and entertainment company with a give attention to digital media.
1. How has ML made a positive impact on Media?
Privacy first personalization for patrons: Every user is exclusive and while their long-term interests are stable, their short-term interests are stochastic. They expect their relationship with the Media to reflect this. The mix of advancement in hardware acceleration and Deep Learning for recommendations has unlocked the flexibility to start out deciphering this nuance and serve users with the correct content at the correct time at the correct touchpoint.
Assistive tools for makers: Makers are the limited assets in media and preserving their creative bandwidth by ML driven human-in-the-loop assistive tools have seen an outsized impact. Something so simple as routinely suggesting an appropriate title, image, video, and/or product that may associate with the content they’re creating unlocks a collaborative machine-human flywheel.
Tightened testing: In a capital intensive media enterprise, there’s a must shorten the time between collecting information on what resonates with users and immediately acting on it. With a wide selection of Bayesian techniques and advancements in reinforcement learning, we’ve been in a position to drastically reduce not only the time but the price related to it.
2. What are the most important ML challenges inside Media?
Privacy, editorial voice, and equitable coverage: Media is a key pillar within the democratic world now greater than ever. ML must respect that and operate inside constraints that usually are not strictly considered table stakes in some other domain or industry. Finding a balance between editorially curated content & programming vs ML driven recommendations continues to be a challenge. One other unique challenge to BuzzFeed is we imagine that the web ought to be free which suggests we do not track our users like others can.
3. What’s a standard mistake you see people make attempting to integrate ML into Media?
Ignoring “the makers” of media: Media is prevalent since it houses a voice that has a deep influence on people. The editors, content creators, writers & makers are the larynx of that voice and the business and constructing ML that permits them, extends their impact and works in harmony with them is the important thing ingredient to success.
4. What excites you most concerning the way forward for ML?
Hopefully, small data-driven general-purpose multi-modal multi-task real-time ML systems that create step-function improvements in drug discovery, high precision surgery, climate control systems & immersive metaverse experiences. Realistically, more accessible, low-effort meta-learning techniques for highly accurate text and image generation.
Li Tan
Background: Li is an AI/ML veteran with 15+ years of experience leading high-profile Data Science teams inside industry leaders like Johnson & Johnson, Microsoft, and Amazon.
Fun Fact: Li continues to be curious, is all the time learning, and enjoys hands-on programming.
Johnson & Johnson: A Multinational corporation that develops medical devices, pharmaceuticals, and consumer packaged goods.
1. How has ML made a positive impact on Pharmaceuticals?
AI/ML applications have exploded within the pharmaceuticals space the past few years and are making many long-term positive impacts. Pharmaceuticals and healthcare have many use cases that may leverage AI/ML.
Applications range from research, and real-world evidence, to smart manufacturing and quality assurance. The technologies used are also very broad: NLP/NLU, CV, AIIoT, Reinforcement Learning, etc. even things like AlphaFold.
2. What are the most important ML challenges inside Pharmaceuticals?
The largest ML challenge inside pharma and healthcare is how you can ensure equality and variety in AI applications. For instance, how you can ensure the training set has good representations of all ethnic groups. Resulting from the character of healthcare and pharma, this problem can have a much greater impact in comparison with applications in another fields.
3. What’s a standard mistake you see people make attempting to integrate ML into Pharmaceuticals?
Wouldn’t say that is necessarily a mistake, but I see many individuals gravitate toward extreme perspectives in relation to AI applications in healthcare; either too conservative or too aggressive.
Some individuals are resistant attributable to high regulatory requirements. We needed to qualify a lot of our AI applications with strict GxP validation. It could require a good amount of labor, but we imagine the trouble is worth it. On the alternative end of the spectrum, there are numerous individuals who think AI/Deep Learning models can outperform humans in lots of applications and run completely autonomously.
As practitioners, we all know that currently, neither is true.
ML models might be incredibly helpful but still make mistakes. So I like to recommend a more progressive approach. The bottom line is to have a framework that may leverage the ability of AI while having goalkeepers in place. FDA has taken actions to control how AI/ML ought to be utilized in software as a medical device and I imagine that’s a positive step forward for our industry.
4. What excites you most concerning the way forward for ML?
The intersections between AI/ML and other hard sciences and technologies. I’m excited to see what’s to return.
Alina Zare
Background: Alina Zare teaches and conducts research in the world of machine learning and artificial intelligence as a Professor within the Electrical and Computer Engineering Department on the University of Florida and Director of the Machine Learning and Sensing Lab. Dr. Zare’s research has focused totally on developing recent machine learning algorithms to routinely understand and process data and imagery.
Her research work has included automated plant root phenotyping, sub-pixel hyperspectral image evaluation, goal detection, and underwater scene understanding using synthetic aperture sonar, LIDAR data evaluation, Ground Penetrating Radar evaluation, and buried landmine and explosive hazard detection.
Fun Fact: Alina is a rower. She joined the crew team in highschool, rowed throughout college and grad school, was head coach of the University of Missouri team while she was an assistant professor, after which rowed as a masters rower when she joined the school at UF.
Machine Learning & Sensing Laboratory: A University of Florida laboratory that develops machine learning methods for autonomously analyzing and understanding sensor data.
1. How has ML made a positive impact on Science
ML has made a positive impact in quite a few ways from helping to automate tedious and/or slow tasks or providing recent ways to look at and take a look at various questions. One example from my work in ML for plant science is that we’ve developed ML approaches to automate plant root segmentation and characterization in imagery. This task was previously a bottleneck for plant scientists taking a look at root imagery. By automating this step through ML we are able to conduct these analyses at a much higher throughput and start to make use of this data to analyze plant biology research questions at scale.
2. What are the most important ML challenges inside Scientific research?
There are various challenges. One example is when using ML for Science research, we’ve to think twice through the info collection and curation protocols. In some cases, the protocols we used for non-ML evaluation usually are not appropriate or effective. The standard of the info and the way representative it’s of what we expect to see in the appliance could make a big impact on the performance, reliability, and trustworthiness of an ML-based system.
3. What’s a standard mistake you see people make attempting to integrate ML into Science?
Related to the query above, one common mistake is misinterpreting results or performance to be a function of just the ML system and never also considering the info collection, curation, calibration, and normalization protocols.
4. What excites you most concerning the way forward for ML?
There are quite a lot of really exciting directions. Plenty of my research currently is in spaces where we’ve an enormous amount of prior knowledge and empirically derived models. For instance, I actually have ongoing work using ML for forest ecology research. The forestry community has a wealthy body of prior knowledge and current purely data-driven ML systems usually are not leveraging. I feel hybrid methods that seamlessly mix prior knowledge with ML approaches will probably be an interesting and exciting path forward.
An example could also be understanding how likely two species are to co-occur in an area. Or what species distribution we could expect given certain environmental conditions. These could potentially be used w/ data-driven methods to make predictions in changing conditions.
Nathan Cahill
Background: Nathan is a passionate machine learning leader with 7 years of experience in research and development, and three years experience creating business value by shipping ML models to prod. He makes a speciality of finding and strategically prioritizing the business’ biggest pain points: unlocking the ability of knowledge earlier on in the expansion curve.
Fun Fact: Before moving into transportation and logistics I used to be engineering rockets at Northrop Grumman. #RocketScience
Xpress Technologies: A digital freight matching technology to attach Shippers, Brokers and Carriers to bring efficiency and automation to the Transportation Industry.
1. How has ML made a positive impact on Logistics/Transportation?
The transportation industry is incredibly fragmented. The highest players in the sport have lower than 1% market share. In consequence, there exist inefficiencies that might be solved by digital solutions.
For instance, while you see a semi-truck driving on the road, there’s currently a 20% probability that the truck is driving with nothing within the back. Yes, 20% of the miles a tractor-trailer drives are from the last drop off of their previous load to their next pickup. The likelihood is that there’s one other truck driving empty (or “deadheading”) in the opposite direction.
With machine learning and optimization this deadhead percent might be reduced significantly, and just taking that number from 20% to 19% percent would cut the equivalent carbon emissions of 100,000 Americans.
Note: the carbon emissions of 100k Americans were my very own back of the napkin math.
2. What are the most important ML challenges inside Logistics?
The massive challenge inside logistics is attributable to the indisputable fact that the industry is so fragmented: there isn’t a shared pool of knowledge that might allow technology solutions to “see” the massive picture. For instance a big fraction of brokerage loads, possibly a majority, costs are negotiated on a load by load basis making them highly volatile. This makes pricing a really difficult problem to resolve. If the industry became more transparent and shared data more freely, so way more would develop into possible.
3. What’s a standard mistake you see people make attempting to integrate ML into Logistics?
I feel that essentially the most common mistake I see is people doing ML and Data Science in a vacuum.
Most ML applications inside logistics will significantly change the dynamics of the issue in the event that they are getting used so it is important to develop models iteratively with the business and ensure that performance in point of fact matches what you expect in training.
An example of that is in pricing where in case you underprice a lane barely, your prices could also be too competitive which is able to create an influx of freight on that lane. This, in turn, may cause costs to go up because the brokers struggle to search out capability for those loads, exacerbating the problem.
4. What excites you essentially the most concerning the way forward for ML?
I feel the thing that excites me most about ML is the chance to make people higher at their jobs.
As ML begins to be ubiquitous in business, it’ll give you the chance to assist speed up decisions and automate redundant work. It will speed up the pace of innovation and create immense economic value. I can’t wait to see what problems we solve in the following 10 years aided by data science and ML!
Nicolas Bertagnolli
Background: Nic is a scientist and engineer working to enhance human communication through machine learning. He’s spent the last decade applying ML/NLP to resolve data problems within the medical space from uncovering novel patterns in cancer genomes to leveraging billions of clinical notes to scale back costs and improve outcomes.
At BEN, Nic innovates intelligent technologies that scale human capabilities to achieve people. See his CV, research, and Medium articles here.
Fun Fact: Nic lived in a van and traveled across the western United States for 3 years before starting work at BEN.
BEN: An entertainment AI company that places brands inside influencer, streaming, TV, and film content to attach brands with audiences in a way that advertisements cannot.
1. How has ML made a positive impact on Marketing?
In so some ways! It’s completely changing the landscape. Marketing is a field steeped in tradition based on gut feelings. Previously 20 years, there was a move to increasingly more statistically informed marketing decisions but many brands are still counting on the gut instincts of their marketing departments. ML is revolutionizing this. With the flexibility to investigate data about which advertisements perform well we are able to make really informed decisions about how and who we market to.
At BEN, ML has really helped us take the guesswork out of quite a lot of the method when coping with influencer marketing. Data helps shine a light-weight through the fog of bias and subjectivity in order that we are able to make informed decisions.
That’s just the plain stuff! ML can be making it possible to make safer marketing decisions for brands. For instance, it’s illegal to advertise alcohol to people under the age of 21. Using machine learning we are able to discover influencers whose audiences are mainly above 21. This scales our ability to assist alcohol brands, and likewise brands who’re frightened about their image being related to alcohol.
2. What are the most important ML challenges inside Marketing?
As with most things in Machine Learning the issues often aren’t really with the models themselves. With tools like Hugging Face, torch hub, etc. so many great and versatile models can be found to work with.
The true challenges need to do with collecting, cleansing, and managing the info. If we wish to speak concerning the hard ML-y bits of the job, a few of it comes right down to the indisputable fact that there’s quite a lot of noise in what people view and luxuriate in. Understanding things like virality are really really hard.
Understanding what makes a creator/influencer successful over time is absolutely hard. There may be quite a lot of weird preference information buried in some pretty noisy difficult-to-acquire data. These problems come right down to having really solid communication between data, ML, and business teams, and constructing models which augment and collaborate with humans as a substitute of fully automating away their roles.
3. What’s a standard mistake you see people make attempting to integrate ML into Marketing?
I don’t think that is exclusive to marketing but prioritizing machine learning and data science over good infrastructure is an enormous problem I see often. Organizations hear about ML and wish to get a chunk of the pie in order that they hire some data scientists only to search out out that they don’t have any infrastructure to service their recent fancy pants models. A ton of the worth of ML is within the infrastructure across the models and in case you’ve got trained models but no infrastructure you’re hosed.
One in every of the very nice things about BEN is we invested heavily in our data infrastructure and built the horse before the cart. Now Data Scientists can construct models that get served to our end users quickly as a substitute of getting to determine every step of that pipeline themselves. Spend money on data engineering before hiring plenty of ML folks.
4. What excites you most concerning the way forward for ML?
There may be a lot exciting stuff occurring. I feel the pace and democratization of the sector is probably what I find most enjoyable. I remember almost 10 years ago writing my first seq2seq model for language translation. It was lots of of lines of code, took perpetually to coach and was pretty difficult. Now you may principally construct a system to translate any language to some other language in under 100 lines of python code. It’s insane! This trend is most definitely to proceed and because the ML infrastructure gets higher and higher it’ll be easier and easier for people without deep domain expertise to deploy and serve models to other people.
Very similar to at first of the web, software developers were few and much between and also you needed a talented team to establish a web site. Then things like Django, Rails, etc. got here out making website constructing easy but serving it was hard. We’re sort of at this place where constructing the models is straightforward but serving them reliably, monitoring them reliably, etc. remains to be difficult. I feel in the following few years the barrier to entry goes to return WAY down here and principally, any high schooler could deploy a deep transformer to some cloud infrastructure and begin serving useful results to the overall population. This is absolutely exciting since it means we’ll begin to see increasingly more tangible innovation, very similar to the explosion of online services. So many cool things!
Eric Golinko
Background: Experienced data practitioner and team builder. I’ve worked in lots of industries across corporations of various sizes. I’m an issue solver, by training a mathematician and computer scientist. But, above all, I value relationships, family, dogs, travel and pizza.
Fun Fact: Eric adores nachos!
E Source: Provides independent market intelligence, consulting, and predictive data science to utilities, major energy users, and other key players within the retail energy marketplace.
1. How has ML made a positive impact on the Energy/Utility industry?
Access to business insight. Provided a pre-requisite is great data. Utilities have many data relationships inside their data portfolio from customers to devices, more specifically, this speaks to monthly billing amounts and enrollment in energy savings programs. Data like that may very well be stored in a relational database, whereas device or asset data we are able to consider because the pieces of machinery that make our grid. Bridging those sorts of data is non-trivial.
As well as, third-party data spatial/gis and weather are extremely vital. Through the lens of machine learning, we’re in a position to find and explore features and outcomes which have an actual impact.
2. What are the most important ML challenges inside Utilities?
There may be a demystification that should occur. What machine learning can do and where it must be monitored or could fall short. The utility industry has established ways of operating, machine learning might be perceived as a disruptor. For this reason, departments might be slow to adopt any recent technology or paradigm. Nevertheless, if the practitioner is in a position to prove results, then results create traction and a bigger appetite to adopt. Additional challenges are on-premise data and access to the cloud and infrastructure. It’s a gradual process and has a learning curve that requires patience.
3. What’s a standard mistake you see people make attempting to integrate ML into Utilities?
Not unique to utilizes, but moving too fast and neglecting good data quality and straightforward quality checks. Except for this machine learning is practiced amongst many groups in some direct or indirect way. A challenge is integrating best development practices across teams. This also means model tracking and with the ability to persist experiments and continuous discovery.
4. What excites you most concerning the way forward for ML?
I’ve been doing this for over a decade, and I by some means still feel like a novice. I feel fortunate to have been a part of teams where I’d be lucky to be called the typical member. My feeling is that the following ten years and beyond will probably be more focused on data engineering to see even a bigger variety of use cases covered by machine learning.
🤗 Thanks for joining us in this primary installment of ML Director Insights. Stay tuned for more insights from ML Directors in SaaS, Finance, and e-Commerce.
Big due to Eric Golinko, Nicolas Bertagnolli, Nathan Cahill, Alina Zare, Li Tan, and Archi Mitra for his or her good insights and participation on this piece. We look ahead to watching each of your continued successes and will probably be cheering you on each step of the way in which. 🎉
Lastly, in case you or your team are keen on accelerating your ML roadmap with Hugging Face Experts please visit hf.co/support to learn more.
