Recruiting the right talent to fit your company has always been difficult and it is no easier within the data science field. The use of AI in recruitment has increased significantly, in an attempt to reduce bias and to provide companies with the ability to identify the most fitting talent for their teams. Ahead of the Deep Learning Summit, taking place in San Francisco on the 24 & 25 January, we spoke to Matt Cowell, CEO at QuantHub, who will be presenting on the Connect Stage, to hear more about their current work in helping companies build their data science and analytics teams with an AI-driven talent assessment platform to attract, vet, and develop data scientists, reducing the time-to-fill those tough data scientist positions.
Give us an overview of QuantHub & your role there.
QuantHub exists to empower companies to conquer the quant crunch, created by the disruptive rise of AI, by providing an AI-driven talent assessment and development platform to attract, vet, and develop data scientists, reducing the time-to-fill those tough data scientist positions and ensuring the team already in place is developed, engaged, and ultimately, retained.
My role as CEO of QuantHub is focused on seeing the world through our customers’ eyes and creating solutions to enable them to build exceptional data science and analytics teams.
Who is the ideal user for QuantHub?
We’ve found companies tend to fall into one of two camps:
Those struggling with bad hires and high attrition due to insufficient technical vetting processes and limited ongoing skill development for team members
Those looking to free up time spent by their valuable data scientists in the hiring/skill development process to instead focus on strategic data science priorities and projects.
In either case, companies stand to gain significant value out of a standardized and automated process proven to delineate between those candidates who have strong technical skills in statistics, linear algebra, modeling, data wrangling, etc. and those who don’t. We understand that great technical skills aren’t the only important characteristic of a great data scientist. Our goal is to help companies more efficiently and effectively identify those candidates with the technical skills needed to excel, allowing companies to spend more of their in-person efforts assessing candidates on cultural fit, teamwork, etc.
Companies focused on leveling up the skills of their team and engaging them in their skill development stand to gain a lot from QuantHub’s targeted, active recall-based, approach to learning and development. Rather than spending your training/development budget on bootcamps and training covering a broad array of topics, QuantHub focuses in on the exact skill development needs for each employee and presents curated training opportunities from industry standard learning platforms to ensure each employee has a skill development plan specifically tailored to their unique needs.
How do you use machine learning to help vet candidates for your users?
QuantHub leverages ML throughout the vetting and development aspects of the platform, including:
- Adaptive-difficulty testing leveraging item response theory for difficulty-level calibration and Bayesian inference to efficiently determine skill levels based on candidate responses
- Testing/assessment content covering ML approaches such as GLM, ensembles, time series, SVM, and deep learning applications such as NLP, computer vision, etc.
- Resume/job matching by extracting keywords/phrases and entities and scoring/ranking the fit between requirements and candidate resumes
- Automated bot-based interviewing leveraging language models and dynamic dialogues with natural language response evaluation
What challenges have you had with implementing deep learning in your work and how have you overcome these?
We’re relatively early on in leveraging deep learning for QuantHub, but we’re encountering common challenges such as a lack of data where we have needed to borrow learnings (weight embeddings) from similar problems and solutions to apply to the problem we’re attempting to solve. Deployment has also been a challenge, but we’ve balanced model complexity with the capacity to execute in production fairly effectively to work around that challenge.
What developments of AI are you most excited for, and which industries do you think will be most impacted?
At QuantHub, we’re excited about the recent rise of transfer learning in NLP and the ability to quickly adapt language models and word embeddings to solve even loosely related problems. NLP will have a major impact on human-centered industries (marketing, customer support, etc.) as it makes it much more accessible for humans to communicate with computers. We’re also excited about advances in Reinforcement Learning and the ability to train an agent to optimize a policy by interacting with its environment. Reinforcement will have a significant impact on machine-centered industries (manufacturing, trading, supply-chain) as it allows for the delegation of (better) decision-making to machines.
Would you advise a career in AI, and what are the key skills that you think are needed for such roles?
In time, most jobs will be impacted by AI in one way or another, so we strongly believe our workforce in general needs to become more data-driven and analytically inclined. However, that doesn’t necessarily mean everyone needs to become a data scientist or ML engineer. As this field continues to mature, other roles will emerge and become more prominent in this field, including roles already critical in our sister field of software development such as Product Management, UX (focused on data), etc. These roles, in addition to the requisite data scientist roles, will go a long way towards helping companies take the innovations with disruptive potential that are surfacing daily and applying those to real-world business problems to deliver significant value to businesses.
Both hard and soft skills are critical. Hard skills to have/develop to pursue a successful career in AI include:
- Mathematics - a foundational knowledge in statistics, linear algebra, and calculus is necessary in order to first understand and analyze your data and then to grasp and leverage the algorithmic approaches of modeling data using machine learning and deep learning.
- Data comfort - A general comfort with data and the ability to retrieve, analyze, combine, and visualize data is critical to being successful in AI and analytics, in general.
- Programming – AI is CS intensive, so proficiency in a programming language such as Python or R is a must.
From a soft skill perspective, a few of the most critical skills include:
- Problem-solving – AI is most effectively applied by great problem solvers who see AI as another tool to solve business/customer problems.
- Resourcefulness – to apply AI to solve real-world challenges, you have to piece together the tools you can find and adapt approaches that in some cases have been created in an academic lab.
- Adaptability to change - methodologies are changing daily but the strongest people will be able to let go of their preferred approaches and adapt to the best way to solve the problem at hand.
- Capacity to learn - similar to the above, methodologies, tools, technologies, etc. are constantly changing. The ability to learn underlying principles and shift those into new frameworks (i.e. from Tensorflow to Pytorch or vice-versa) is critical.
What are you most excited for at the upcoming Deep Learning Summit in San Francisco?
Beyond making a trip to Scoma’s for dinner :-), we’re really looking forward to networking with peers to compare notes and generate new ideas and new ways we can better apply machine learning and deep learning within our solutions to help our customers. We’re also looking forward to talking to companies about their models for vetting and developing data scientists to learn how progressive organizations are approaching this critical area and see how QuantHub can help.