Data science

Data science

AI was everywhere, but I wasn't focused on product launches. I was looking at how companies think about data itself: how it's shared, governed and ultimately turned into decisions. And across conversations with executives and sessions on security and compliance, a pattern emerged: the technical limitations that once justified locking data down have largely been solved. What remains difficult is human. Alignment, trust and confidence inside organizations are now the true barriers.

Every search, purchase, loyalty swipe, location ping and scroll feeds systems that now shape pricing, product decisions, hiring and marketing strategies. Most founders understand this in theory, but few grasp the practical consequence: whether they intend to or not, they and their customers are already casting votes with their data. And those votes? They're usually cast passively, on someone else's terms.

I wrote a book for O'Reilly on scaling machine learning with Spark specifically. My second book is coming out on how to improve high-performance Spark, the second edition. Started my career in the machine learning space 15 years ago, moved into data infrastructure, batch processing, and a year and a half ago I moved into the data streaming space, which I think it's what's going to help us pave the future in the data.

Most beginner data portfolios look similar. They include: A few cleaned datasets Some charts or dashboards A notebook with code and commentary Again, nothing here is wrong. But hiring teams don't review portfolios to check whether you can follow instructions. They review them to see whether you can think like a data analyst. When projects feel generic, reviewers are left guessing:

"Pyramid adds an analytics and semantic layer that can define metrics in a way that both humans and AI agents can rely on,"

Every year, poor communication and siloed data bleed companies of productivity and profit. Research shows U.S. businesses lose up to $1.2 trillion annually to ineffective communication, that's about $12,506 per employee per year. This stems from breakdowns that waste an average of 7.47 hours per employee each week on miscommunications. The damage isn't only interpersonal; it's structural. Disconnected and fragmented data systems mean that employees spend around 12 hours per week just searching for information trapped in those silos.

For a brief moment in October, Alejandro Quintero thought he had made it big in China. The Bogotá-based data analyst owns and manages a website that publishes articles about paranormal activities, like ghosts and aliens. The content is written in "Spanglish," he says, and was never intended for an Asian audience. But last fall, Quintero's site suddenly began receiving a large volume of visits from China and Singapore.

Next Gen Stats began in 2015, when the National Football League deployed RFID chips in player shoulder pads and even in the football itself, enabling the league to capture location data multiple times per second through sensors installed throughout stadiums.

This is the conundrum of elite chess. The stronger the players, the greater the odds of the match ending in a draw. "What ended up happening," said Mark Glickman, senior lecturer in the Department of Statistics and longtime chess enthusiast, "is that these top players were not having their ratings change very much, just because the games would be drawn all the time."

A new PhD track is being added to the Walter S. and Lucienne Driskill Graduate Program in Life Sciences ( DGP) for the 2026 application cycle, to enhance student learning and build community around computational biology and bioinformatics at Feinberg. The computational biology and bioinformatics (CBB) track in the graduate program will prepare students through coursework and lectures to use modern computational approaches, including machine learning and artificial intelligence, to extract biological insight from large-scale datasets to address complex biological problems.

That local exodus is documented by Cornell-led research that mapped annual moves between U.S. neighborhoods from 2010 to 2019 in detail 4,600 times greater than standard public data. Called MIGRATE, the new, publicly available dataset revealed that most of those displaced remained within the affected county - moves not captured in county-level public migration data aggregated every five years.

Instead of treating each prompt as a one-off request, the new agent remembers what was asked earlier, including datasets, filters, time ranges, and assumptions, and uses that context when answering follow-up questions. This lets users refine an analysis progressively rather than starting from scratch each time," Satapathy added. Satapathy pointed out that this eases the pressure on developers to prebuild dashboards or predefined business logic for every possible question that a data analyst or business user could ask.

With the introduction of Live Query for BigQuery and Alteryx One: Google Edition, users no longer need to move data to run workflows. Companies that standardize cloud platforms for analytics and AI often see a gap between where data is stored and how it is prepared and used. Alteryx wants to change that by bringing analytics workflows directly to BigQuery. The promise: from data to insight to action, without compromising on security or scalability.

From a meteorological perspective, the winter storm sweeping across the country this weekend is a supply chain disruption in its own right: A high-pressure system from the north is smashing into a low-pressure system from the south, belting large swaths of the US with heavy snow, sleet, and freezing rain. While the snarl in the upper atmosphere could trickle down to the real supply chain on the ground, some retailers are taking steps to anticipate the impact of the storm and position their products accordingly.

Our work is not about producing a list of tables with numbers in rows and columns,

You can always make it better. You can improve things. But it does give you a good taste of what can be done in vibe coding. Those are things that I made maybe in 15 minutes, half an hour. It is quite simple to get those first steps and say, "Oh, this works." Maybe you want to do some improvements, and you refine the code and what you're expecting.

I'm thrilled to announce that I'm stepping up as Probabl 's CSO (Chief Science Officer) to supercharge scikit-learn and its ecosystem, pursuing my dreams of tools that help go from data to impact. Scikit-learn, a central tool Scikit-learn is central to data-scientists' work: it is the most used machine-learning package. It has grown over more than a decade, supported by volunteers' time, donations, and grant funding, with a central role of Inria.

The more attributes you add to your metrics, the more complex and valuable questions you can answer. Every additional attribute provides a new dimension for analysis and troubleshooting. For instance, adding an infrastructure attribute, such as region can help you determine if a performance issue is isolated to a specific geographic area or is widespread. Similarly, adding business context, like a store location attribute for an e-commerce platform, allows you to understand if an issue is specific to a particular set of stores

"When I first started this job, the main push back I always got was that synthetic data will take over and you just will not need human feedback two to three years from now," said Fitzpatrick, who joined the startup last year. "From first principles, that actually doesn't make very much sense." Synthetic data refers to data that is artificially created.

Another year. It passed extremely fast and yet, painfully slow. Despite developing tech that some think might take over our day-to-day work, data things got made by people this year. These are my favorites. Inside the Confusing World of Women's Clothing Sizes They approached the topic from several angles with 3-D models, data collection, and sizing charts. Adding to the visualization genre of variable clothes sizes, this piece helped me appreciate the process that is women's shopping. [ See the Project / On FlowingData]

When Sam Altman said one year ago that OpenAI's Roman Empire is the actual Roman Empire, he wasn't kidding. In the same way that the Romans gradually amassed an empire of land spanning three continents and one-ninth of the Earth's circumference, the CEO and his cohort are now dotting the planet with their own latifundia-not agricultural estates, but AI data centers.

Upper is based on W3C standards such as RDF for conceptual graph representation and SHACL for validation, and it enables the principle of "model once, represent everywhere" across the data ecosystem.Upper organizes concepts through keyed entities, their attributes, and their relationships across domain boundaries. The modeling grammar and validation structure are designed to maintain consistency as definitions evolve. Keyed concepts can be extended monotonically, allowing new attributes or relationships without modifying existing definitions allowing domains to expand over time without breaking existing models.

The pandas DataFrame is a structure that contains two-dimensional data and its corresponding labels. DataFrames are widely used in data science, machine learning, scientific computing, and many other data-intensive fields. DataFrames are similar to SQL tables or the spreadsheets that you work with in Excel or Calc.

Speaking to investment analysts, he said that while MongoDB had all the elements needed to be the right foundational platform for AI workloads, it was too early to say what might be the platform of choice. However, he said MongoDB had been winning work from AI-native companies, citing a customer that recently "switched from PostgreSQL to MongoDB because PostgreSQL could not just scale."

Snowflake has signed an agreement to acquire Select Star. This company's technology will expand Snowflake Horizon Catalog by integrating with databases, BI tools, and data pipelines. This will increase the context for AI agents such as Snowflake Intelligence. The full context of data assets is often scattered across upstream and downstream systems. This fragmentation makes it difficult to find the right data and understand the full context. In the AI era, this limited context poses a problem for both humans and agents.