phData

Today marks another milestone in the phData and Snowflake collaboration.

Together, we're combining Snowflake's native AI capabilities with phData's engineering expertise to help enterprises modernize analytics, accelerate migrations, build governed AI applications, and turn AI into operational business value at scale.

"Enterprise customers need a partner who knows how to implement with precision. phData has consistently delivered that." — Kayle McBride, VP of GSI and Americas Alliances, Snowflake

Read the full announcement: https://www.phdata.io/blog/phdata-snowflake-strategic-collaboration-2026/

phData and Snowflake Expand Strategic Collaboration to Help Enterprises Put AI Into Production phData and Snowflake expand their strategic collaboration to help enterprises deploy governed AI on Cortex AI and CoCo — faster, at scale, and in production.

Claude is powerful. Making it productive at enterprise scale takes more than a model.

phData helps enterprises put Claude to work inside the systems that run the business: trusted data, connected workflows, and production-grade AI with governance built in from day one.

The result is not another promising demo. It is a system that delivers measurable value, improves over time, and is built to operate.

As an Anthropic Preferred Services Partner, phData helps organizations move from high-potential use cases to production with Claude.

Explore what we build: https://www.phdata.io/partners/anthropic/

Most KNIME engineers reconfigure the same nodes manually before every run. File paths. Filter values. Date ranges. Changed by hand, every time.

Bianca Sterpone ran into this across multiple client projects at phData, including Alteryx-to-KNIME migrations where macro logic was a big part of the work. Flow variables were the answer.

Bianca wrote a blog on how they work, where people get tripped up, and how to build workflows that actually hold up in production.

Find the blog link in the comments.

A life sciences coordinator has 14 browser tabs open.

- Courier portal: temperature excursion alert.
- CDMO portal: capacity reduced.

The ERP says everything is on track.

ERP systems think in batches and purchase orders. Cell and gene therapy runs on individual patients, narrow viability windows, and manufacturing slots measured in units of one.

The pattern that works:
1️⃣ Keep ERP as the system of record for finance and compliance.
2️⃣ Build an intelligence layer above it that ingests data from labs, DCTs, logistics, and portals, and feeds decision-ready insights back in a form the business can actually use.

ERP stays. It just stops being where you go to understand the health of a therapy program.

phData's VP of Life Sciences, Deepti Cole, wrote the full breakdown in the blog. Read it now: https://www.phdata.io/blog/why-traditional-erp-systems-fail-to-track-the-patient-journey-in-life-sciences/

Keystone Cooperative had seven years of agricultural production data. Feed logs, weights, and group lifecycle records. Nobody had connected it.

phData built a custom AI weight prediction model that tells Keystone's production team exactly when to market. Get the right pig on the truck at the right weight, and every percent of margin captured flows back to member-owners through patronage.

At thousands of head per week, that compounds fast.

Read how Keystone turned seven years of idle data into grower dividends: https://www.porkbusiness.com/news/hog-production/keystone-cooperative-leverages-ai-turn-data-grower-dividends

Build a lakehouse on Amazon Web Services that your business can actually use.

Join phData and AWS in person for a half-day session on Apache Iceberg, followed by a hands-on lab in the afternoon.

Learn how leading companies are combining open data architectures with AI agents to reduce platform costs, eliminate vendor lock-in, and put insights directly into the hands of business leaders.

Register here: https://aws-experience.com/amer/smb/e/132c1/the-apache-iceberg-on-aws-advantage-building-powerful-lakehouses-for-the-agentic-ai-era

Most AI investment goes into models. Almost none goes into whether the model actually knows what your data means. That gap is where reliability breaks down.

Data platforms answer "what do we have, and where does it live?"

Intelligence platforms have to answer a harder question: "what does it mean, and how should it be interpreted?"

The semantic layer, the definitions, metrics, and ownership of terms, is what makes the difference. And it's usually the last thing built.

Dustin Dorsey gets into how phData approaches that foundation in practice.

Read the blog: https://www.getdbt.com/blog/data-platforms-were-built-to-store-intelligence-platforms-are-built-to-reason