John Chan, Director of Expertise – AI/ML, Raymond James
John Chan is a Director of Expertise at Raymond James Monetary working the Carillon Labs – the innovation labs specializing in AI/ML. His ardour is to advertise AI adoptions and implement machine studying options within the monetary sector. He has over 20 years of expertise main and implementing expertise options from FinTech startups to top-tier banks and consulting companies. Earlier than Raymond James, John was a synthetic intelligence (AI) strategist and engineering lead at Gamma Lab of OneConnect Monetary, Morgan Stanley knowledge science group and KPMG Cognitive Expertise Lab. He’s lively in NLP analysis specializing in Generative AI, Conversational AI, Doc Understanding and danger and compliance expertise. He’s a frequent speaker at AI occasions.
1. Are you able to share some insights into your skilled journey and the important thing experiences that led you to your present function at Raymond James?
I began my skilled journey as a knowledge analyst. About 10 years in the past, I reinvented myself in knowledge science and AI. I imagine AI is essentially the most important revolution in human historical past, surpassing what the Industrial Revolution has dropped at us.
Fixing enterprise issues and discovering methods to ship simpler options, I’ve had the privilege of collaborating with sensible, gifted people and forming sturdy groups able to tackling challenges. It is very important mix artwork and science, understanding the advanced dynamic of individuals and alter whereas aligning visions with methods and pursuing engineering excellence. These experiences and passions have collectively formed my path, main me to my present function.
2. What about essentially the most important developments and developments in AI/ML that you just imagine will impression the monetary trade, significantly in wealth administration and funding banking?
The brand new wave of the AI revolution is simply beginning. There isn’t any doubt that it’ll be extra impactful to the monetary trade or any trade that requires utilizing language to make choices or create content material. Final 12 months, we noticed industries flooded with GenAI subjects. Presently, we’ve got RAG added to the combination, concentrating on some shortcomings of GenAI or LLM on the whole.
To attain AI execution excellence, we want speedy methods and tactical fight decision-making mindsets, much like the enterprise mindsets to win enterprise
The following 12 months is more likely to be Agentic AI, the place many RAG/LLMs can be working collectively to push for higher outcomes. With vibrant actions within the analysis group and considerable funding pouring into AI, I imagine the AI mannequin accuracy will quickly blow our minds. Not solely the wealth administration and funding banking industries however just about any industries that require well timed communication and are heavy on paperwork can be remodeled by AI/ML, particularly by GenAI.
3. What do you see as the most important challenges for AI/ML in your trade over the last decade, and the way are you getting ready to deal with them?
I believe the toughest factor is to navigate with agility over forms. We’re experiencing unprecedented shear velocity challenges. Extremely regulated industries are used to transferring slowly. The dearth of velocity to deal with the velocity of AI modifications will expose vulnerabilities, giving opponents alternatives to steal market shares.
The outdated mannequin is Attempt-Sluggish, Fail-Sluggish, and does a variety of storytelling, pushed by multi-year roadmaps and thus can not sustain with the velocity regardless of how a lot tweaking on plans. To attain AI execution excellence, we want speedy methods and tactical fight decision-making mindsets, much like enterprise mindsets to win enterprise. Concretely, we want organizational dedication to vary, ranging from government sponsorship and transferring from high down on quick tracks.
4. What does your present AI/ML group seem like when it comes to roles and experience? How do you make sure the group has the mandatory expertise to remain forward of the curve?
We have now the AI answer group actively interfacing with companies or purchasers to align our applied sciences with the enterprise aims, and the information groups collect, cleanse, analyze, perceive and increase knowledge. The AI engineer and AIOps group are accountable for coding, coaching, finetuning, testing and producing AI providers. We even have the groups for the mannequin danger, together with explainability, privateness, governance and AI ethics.
To make sure the group stays forward of the curve, I begin with a complete screening when constructing groups, making certain they’ll show quick technical expertise and be culturally match. They need to be obsessed with utilizing AI/ML to resolve real-life issues and need to have enjoyable whereas working onerous, and have the motivation to remain forward of the curve.
AI is advancing quick to the purpose that it’s fairly tough to maintain up. I encourage my group to study new issues particular to their area experience, learn blogs and technical papers usually, and share their new data with their teammates to sharpen one another.
5. Are you able to describe the AI/ML expertise stack at the moment in use at Raymond James? What issues went into deciding on these instruments and frameworks?
I’m fairly agnostic to expertise stacks. Many stacks could make issues work. I simply make certain the broader group can agree on the instruments and frameworks that we will stick to for the long run. Many cloud suppliers have complete AI answer stacks. Most of my work is using the AWS cloud ecosystem. I’m positive AzureAI, GCP, and many others. are pretty much as good. For improvement, I’m a python/torch particular person. I like Nvidia DGX, particularly for deep mannequin coaching.
6. What recommendation would you give to different monetary providers companies seeking to undertake AI/ML applied sciences? What are the important thing issues and potential pitfalls they need to pay attention to?
Many companies have initiated AI/ML for fairly a while, some companies are fairly mature, particularly in traditional ML. Nevertheless, GenAI requires completely different expertise and approaches to use because it targets data work automation and making data-informed choices. Organizations have to have an outlined AI Technique that may help long-term digital transformation with sturdy government sponsorship.