Leaders from Apple, DeepMind, Cohere, and more are gathering to confront the demands of modern data science.
Next week, the centre of gravity in data science shifts to Toronto.
From August 3 to 7, Toronto will host KDD-2025, the flagship conference for cutting-edge research and AI applications, organized by the Association for Computing Machinery’s (ACM) Special Interest Group on Knowledge Discovery and Data Mining (SIGKDD).
“While AI is transforming the field, the core mission—turning raw data into meaningful knowledge—remains crucial.”
Jian Pei, KDD
KDD brings together thousands of researchers and tech leaders who work at the frontlines of data science.
It’s a rare venue where academic research meets real-world deployment, bringing people building the infrastructure of AI inside labs, startups, and Fortune 500s together to learn, compare notes, and influence the direction of the field.
This year, leaders from Cohere, Apple, Google DeepMind, Amazon, and Cisco, and many others will join discussions that span from model behaviour to system deployment, safety, and governance.
The 2025 conference takes place at a moment of extraordinary transition for the data science community, as the pace of generative AI development has outstripped many of the field’s assumptions and introduced new pressure points for anyone working with data at scale.
According to Xiaohui Yu, Professor at York University and one of three KDD-2025 General Chairs, this year’s program was designed to reflect that urgency.
“The world of data science and AI is changing fast,” said Yu. “The KDD community—which has been at the heart of discovering insights from data for decades—is evolving along with it.”
The new shape of data science
For decades, data scientists focused on extracting insights from structured datasets. Today, the landscape looks markedly different.
“We are at an inflection point,” said Luiza Antonie, a Professor at the University of Guelph and General Chair of KDD.
One of the most profound shifts Antonie has observed in the last year is how users engage with AI. Instead of receiving outputs from a trained model, data scientists now interact with systems that respond fluidly, mimic human intuition, and generate complex outputs on the fly.
“Think of it like learning how to talk to a very smart but sometimes overly literal assistant,” Antonie said. “You have to phrase things just right to get what you need.”
These changes to how people engage with AI is also shaping its development. According to Jian Pei, KDD General Chair and the Arthur S. Pearse Distinguished Professor at Duke University, instead of building models from scratch, many data scientists now fine-tune large, general-purpose systems.
“This has made AI development faster but also raised concerns about who controls these powerful tools,” Pei said. “Big tech companies dominate the space, while open-source alternatives aspire to keep AI accessible to everyone.”
AI tools are also increasingly involved in the decision-making process for data scientists. As models become more capable, they are entrusted with responsibilities that were once strictly assigned to humans.
The pace of change is creating ripple effects in the data science community. According to Yu, automation is streamlining much of the entry-level work, but demand is growing for professionals who can design systems with safety in mind, integrate AI into complex workflows, and think ahead to the ethical concerns.
“At the same time, AI is becoming easier to use for non-experts, meaning more people can experiment with it without needing a PhD in computer science,” Yu said.
Yu believes the role of the data scientist is shifting from merely writing code and building models to shaping how intelligent systems behave, interact, and scale.
A gathering for the moment
These emergent trends have made KDD’s purpose more relevant than ever, according to the conference’s General Chairs.
“While AI is transforming the field, the core mission—turning raw data into meaningful knowledge—remains crucial,” said Pei.
A data scientist’s mission now requires a different level of thinking. Antonie describes this as “data thinking,” a skillset rooted in judgment, context, and technical fluency, especially when systems generate outputs that can’t easily be audited or explained. Sessions throughout the week at KDD are designed to push that discipline forward.
This year’s program also opens the floor to organizations already deploying generative models. “You’ll hear how retail giants use generative AI to personalize shopping experiences, or how hospitals deploy predictive models to improve patient care,” Pei said.
During the conference’s workshops, researchers will work alongside applied teams to tackle live issues, such as algorithmic bias, privacy architecture, evaluation frameworks, and large-scale deployment. Companies like Google, Microsoft, and early-stage AI firms will also participate directly as technical contributors.
KDD-2025 also includes dedicated activities discussing governance and oversight. Panels will feature regulators, civil society leaders, and applied ethicists, who will discuss how intelligent systems should be developed and deployed.
The conference program was created for professionals solving domain-specific problems in real-world settings. That design comes into focus through six curated Special Days, which span finance, health, responsible AI, Canada AI, AI and science, and AI reasoning. The agenda for each day is set by experts in each domain.
The week’s program also includes something KDD has become known for: its tutorials. These deep, structured sessions function as the conference’s technical backbone, offering working knowledge for researchers, engineers, and data science teams.
Among this year’s tutorial presenters are research heavyweights like the Michael Aiken Chair Professor Jiawei Han and Distinguished Professor Eamonn Keogh, alongside up-and-coming researchers in the space.
One of the most anticipated sessions is from a team led by Dr. Reza Mohammadi, a researcher at SAP focused on evaluating AI agents in enterprise settings. Their tutorial will tackle a growing blind spot in AI development: how to measure whether autonomous systems are actually performing their intended tasks. Mohammadi and colleagues will guide attendees through a framework for evaluating agentic systems, including what to measure and how to measure it.
“Attendees will walk away with a structured way to think about LLM agent evaluation,” Mohammadi said. “Not just a list of metrics, but a taxonomy or an agentic systems evaluation map that connects goals, data, processes, and context.”
‘Innovation with accountability’
Across the tech sector, data scientists are facing pressure to build systems that work reliably, make fair decisions, and stand up to real-world demands. KDD-2025 puts those challenges at the centre of its program and aims to steer the industry to what Pei described as “innovation with accountability.”
For Antonie, the conference brings something irreplaceable to the data science community: “decades of hard-won wisdom about working with actual data in messy reality.”
Join the researchers and industry leaders shaping the future of data science and applied AI. Register for KDD 2025 now.
All photos provided by KDD-2025.