engblogs

Operationalizing Agentic AI Part 1 explains how to convert agentic AI into an executable, governed operating model, defining concrete work, bounded autonomy, observable metrics, and safe fallbacks to bridge the execution gap from pilots to production.

Turning a model into an autonomous agent by provisioning the Responses API with a programmable computer environment.

Graph-based ontologies and open standards enable scalable, auditable, and interoperable knowledge management for nuclear plants, linking design, licensing, safety constraints, and operations to accelerate decision-making under evolving regulatory and technology landscapes.

Wayfair leverages OpenAI to boost catalog accuracy and accelerate customer support.

A practical guide to designing AI agents that resist prompt injection, outlining threat models, defenses, and evaluation strategies.

Practical, multi-layered mitigation of prompt injection for AI agents on Databricks, mapping Pillars of the Agents Rule of Two to DASF-aligned controls—data access, model interaction, and runtime execution—illustrated by the Social Gauge app and Mosaic AI Gateway guardrails.

MIT's Humane UXD course blends anthropology and computer science to teach humane, user-centered chatbot design using linguistic anthropology and real-world projects (Pond, News Nest, M^3) to steer AI toward social guidance rather than addictive engagement.

Modular Structured Mojo Kernels split a GPU matmul into TileIO, TilePipeline, and TileOp, delivering a durable, cross-vendor pattern centered on TileTensor that scales across NVIDIA and AMD without rewriting the kernel structure.

A technical guide to HIMSS26 highlighting how Databricks enables agentic AI at scale on a governed lakehouse for healthcare, powering real-time analytics, predictive insights, and AI-enabled clinical workflows.

An in-depth, developer-focused guide to Databricks Lakebase autoscaling, explaining how elastic Compute Units and scale-to-zero enable responsive, cost-aware databases for feature branches, CI/CD, and modern workloads.

Databricks merges Quotient AI to inject continuous evaluation and learning for AI agents in production, strengthening Genie, Genie Code and Agent Bricks with structured evaluation and reward signals to steadily improve accuracy, reliability and domain expertise.

Genie Code is Databricks’ agentic, autonomous AI for data teams that builds pipelines, debugs failures, ships dashboards, and maintains production systems, tightly integrated with Unity Catalog for governance and data context.