The Hallway Track
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
AWS shows how to build a healthcare voice appointment agent using Amazon Nova 2 Sonic and Bedrock AgentCore.
“Instead of chaining separate transcription, reasoning, and synthesis services, Nova 2 Sonic processes speech natively in a single model—so vocal context like tone and pace isn’t lost to transcription.”
Snowflake semantic views feeding Amazon Quick Sight create a shared governed data layer so AI and BI return consistent answers.
“This leads to trustworthy answers and significantly reduces the risk of AI hallucinations.”
LLM-generated job applications and portfolios are generic and reveal nothing authentic about the candidate.
“The perfected, generated, prompted resume is generic and impersonal. It tells me nothing about this person, other than that they use particular tools.”
Daikin Applied Americas uses Databricks Genie Code for agentic data engineering to build consistent data pipelines at scale
DeepLearningAI promotes a 7-day challenge to build a voice AI agent that calls you when it needs input.
“Join our 7-day build challenge and build an agent that calls you when it needs your input.”
NVIDIA previews RTX Spark laptops with 128GB memory for 12K video editing and real-time DLSS 4.5 rendering.
“we can make decisions as creators because we see how the final content will look like”
Loka built a low-latency voice agent on Amazon Nova 2 Sonic that outscored rival speech-to-speech models on speech reasoning.
“Amazon Nova 2 Sonic achieved a speech reasoning score of 87.0.”
Kythera Labs is building an AI-native healthcare data platform on Databricks to surface answers already in data.
Azure SQL Database lets developers build vectors on operational data and ground LLMs like OpenAI and Anthropic via Microsoft Foundry.
“You can you can essentially talk to LLMs with the with the truth that you have in your operational data.”
Google DeepMind adds computer use capabilities to its Gemini 3.5 Flash model.
NVIDIA details GPU-accelerated BEV pooling optimizations for autonomous vehicle, robotics, and spatial AI perception pipelines.
“BEV models project multicamera image features into a shared top-down grid, providing downstream perception and planning modules with a common spatial layout for reasoning about lanes, vehicles, pedestrians, and free space.”
Instacart's AI-powered Caper Carts and connected-store tech aim to unify in-store and online grocery shopping within 5-10 years.
“Our view is that in five to 10 years, customers shouldn't have to think about shopping in store or online.”
LangSmith Engine and Context Hub turn agent traces into durable, versioned long-term memory for continual learning.
“Most agents don't learn, they just leave traces.”
Use dataset distillation and fine-tuning instead of longer prompts to enforce consistent structured outputs.
“Prompting tells the model what you want right now. Fine-tuning teaches the model a pattern it can follow repeatedly.”
NVIDIA NeMo AutoModel accelerates fine-tuning of Hugging Face Transformers models.
Azure SQL Database lets developers build vector search on operational data and ground LLMs like OpenAI and Anthropic via Microsoft Foundry.
“you can essentially talk to LLMs with the truth that you have in your operational data”
Databricks named highest in execution and furthest in vision in Gartner Magic Quadrant for a second year.
Clay runs over 350 million go-to-market AI agents monthly, processing trillions of tokens per week.
“We run this over 350 million times a month. It processes trillions of tokens every week.”
Sierra runs two transcription models in parallel to catch silence hallucinations and avoid single-provider limits.
“there is one model that has the highest quality transcription, but it hallucinates during silence more than other models. So we run two models in parallel.”
Engram, a new lab from Dan Biderman and Jessy Lin, treats memory and continual learning as baking new context directly into model weights.
“We don't see the world through the lens of pre-training or post-training. Our models are always training.”
AWS demoed Bedrock AgentCore features enabling self-improving agents via evaluations, observability, and prompt optimization.
“I drive obserability evaluations, behavioral analysis and recommendation set of capabilities for agents.”
Apache Iceberg with Amazon S3 tables is the best foundation for landing real-time streaming and CDC data into a lakehouse.
“How do you get real-time streaming data into your data lakehouse and efficiently? The answer, increasingly, is Apache Iceberg.”
Anthropic launched Claude Tag, a Slack-native way to delegate work to Claude as a proactive, persistent teammate.
“third major redesign of LLM UIUX”
AWS unveils new Strands Agents features and launches Lambda micro VMs with stateful persistent storage running up to 8 hours.
“We just wanted to come up with this model-driven approach and uh release strands that we started using internally.”