Suggestions
Sabya Mukherjee
Leading Gen AI and MLOps at Manulife
Sabyasachi Mukherjee is a seasoned analytics professional with 15+ years of experience in data-centric roles across Fortune 100 companies in various sectors, including financial, telecom, healthcare, consumer goods, and insurance. He has worked internationally in the UK, USA, Canada, and Switzerland, accumulating experience in BI architecture, solution design, data modeling, and OLAP technologies as well as practice and pre-sales management. Mr. Mukherjee is a TOGAF certified architect committed to agile execution. He holds an Electrical and Electronics Engineering degree from National Institute of Technology Karnataka and studied at Taki House Boys.
Highlights
I never considered data governance to be a happening field. I mean, it's just about who gets to see what data and collecting business glossaries, right? I was wrong.
Here's a mind-bender for you: follow the money trail. While everyone's gaga over generative AI, the big bucks are quietly betting on data governance. Here's what happened in last 3 weeks - Databricks acquired storage platform Tabular for 1B+, and both Snowflake and Databricks suddenly open-sourced their data catalogs one after the other. Why would two competing companies give away their core product for free? Apparently it boils down to two things: common data format and data catalogs. Yawn, right? Stay with me.
Think of your access to enterprise data as navigating a city. Without these two, you're navigating with an old paper map—no real-time updates, no alternate routes, or choice of transportation. Now imagine upgrading to Google Maps - that's the shift happening in data governance space.
With tech like Fabric or Delta Sharing, data ingestion is becoming a thing of the past and open file formats are becoming crucial. Format interoperability is the common language letting all systems talk. It's how Google Maps guides you whether you're walking, driving, or taking public transport.
Data catalogs are also transforming data governance. Instead of adjusting permissions in each database, you establish core policies at the catalog level . This mirrors Google Maps' real-time updates feature—changes in data access or availability are reflected instantly across your entire data ecosystem, just like traffic updates appear automatically on your map.
I feel these changes are turning data governance from a chore into a powerful tool, and investment from enterprises in this space will be a decisive factor in getting the most value out of enterprise data.
"There are only two ways to make money in business: One is to bundle; the other is to unbundle." This quote, often attributed to Jim Barksdale, captures the cyclical nature of product and service offerings across industries.
The music industry illustrates this idea: Albums bundled songs, iTunes unbundled them, and now streaming services like Spotify have re-bundled vast libraries for monthly subscriptions. A similar pattern occurred with Cable TV, Netflix, and live TV streaming.
Databricks Data + AI Summit showcased a fascinating trend – bundling within the unbundled world of data and AI platforms. While the last data unbundling separated data and compute, leading to cloud-native and data mesh architectures - Databricks among others emerged as a top leader. Now Databricks' announcements showed a shift towards bundling, simplifying user access to their data intelligence platform.
In my opinion five key announcements highlight this:
- Serverless: The entire Databricks platform now offers serverless. Eliminating cluster and infrastructure management – one of the recurring pain point today for all cloud services.
- Uniform: Full compatibility between Delta Lake and Apache Iceberg formats (doubling down by acquiring Tabular (now part of Databricks))
- Lakeflow: Bundling data ingestion, transformation, and orchestration.
- Delta Sharing: Secure table sharing across clouds and platforms - eliminating complex data ingestion processes.
- Databricks AI/BI: AI-first approach to business intelligence with conversational agent Genie.
There's no permanent sweet spot. The pendulum swings between bundling and unbundling as technology, consumer preferences, and market conditions evolve – creating opportunities for innovation and value creation in both directions. Databricks' new focus seems to be simplifying the data and AI landscape by offering a comprehensive, bundled solution for users of all expertise levels – more power to the consumers.
My reflection on #DataAISummit
