Peninsula AI, Machine Learning and Computer Meetup

Oct 1, 2026
6:00 PM - 8:00 PM PDT
Online. Register for the Zoom!

Beyond Exact Matches: Detecting Modified 3D Assets at Marketplace Scale

How can a marketplace identify copied 3D assets when their orientation, geometry, or composition has changed? Drawing on my work in 3D content understanding at Roblox, this talk will explore multi-view and rotation-invariant representations for similarity and duplicate detection, including the challenges posed by deformed and fragmented copies.
It will examine how geometric and semantic signals can complement one another, and discuss practical trade-offs in evaluating detection quality and deploying these methods at scale. The presentation will draw on published patent applications and publicly shareable examples to offer practical lessons for engineers building visual search, content-understanding, and marketplace-safety systems.

About the Speaker

Phani Harish Wajjala is a Principal Machine Learning Engineer at Roblox specializing in 3D computer vision, multimodal AI, and large-scale content understanding.

Sign Language: Towards Sign Understanding for Robot Autonomy

Navigational signs are common aids for human wayfinding and scene understanding, but are underutilized by robots. We argue that they benefit robot navigation and scene understanding, by directly encoding privileged information on actions, spatial regions, and relations.

Interpreting signs in open-world settings remains a challenge owing to the complexity of scenes and signs, but recent advances in vision-language models (VLMs) make this feasible. To advance progress in this area, we introduce the task of visual sign grounding, which parses locations and associated directions from signs, and maps them to region in the sign’s local environment.

Additionally, we present a baseline approach using VLMs, and demonstrate their promise on the task. We also outline different applications, such as localization and navigation, which benefit from the spatial-symbolic information encoded by navigational signs.

About the Speaker

Nicky Zimmerman I am a postdoctoral researcher in NUS, working on open world navigation. My PhD thesis focused on human-inspired strategies for semantic localization and mapping. Previously, I worked as computer vision algorithm developer in General Motors and Intel.

MeMo: Memory as a Model

Large language models (LLMs) achieve strong performance across a wide range of tasks, but remain frozen after pretraining until subsequent updates. Many real-world applications require timely, domain-specific information, motivating the need for efficient mechanisms to incorporate new knowledge.

In this paper, we introduce MeMo (Memory as a Model), a modular framework that encodes new knowledge into a dedicated Memory model while keeping the LLM unchanged. Compared to existing methods, MeMo offers several advantages: (a) it captures complex cross-document relationships, (b) it is robust to retrieval noise, (c) it avoids catastrophic forgetting in the LLM, (d) it does not require access to the LLM’s weights or output logits that enabling plug-and-play integration with both open and proprietary LLMs, and (e) its retrieval cost is independent of corpus size at inference time.

Our experiments on three benchmarks, BrowseComp-Plus, NarrativeQA, and MuSiQue, show that MeMo achieves strong performance compared to existing methods across diverse settings.

About the Speaker

Arun Verma is a Postdoctoral Associate at the Singapore-MIT Alliance for Research and Technology Centre, where he works with Daniela Rus, Armando Solar-Lezama, and Bryan Low.

This workshop explores the FiftyOne plugin framework to build custom computer vision applications. You’ll learn to extend the FiftyOne App with Python based panels and server side operators, as well as integrate external tools for labeling, vector search, and model inference into your dataset views.

Date, Time and Location

Oct 7, 2026
9 AM - 10 AM PST
Online. Register for the Zoom!

You’ll also automate repetitive tasks by writing custom workflows executing within the FiftyOne environment. Attendees will learn to transform FiftyOne from a visualization tool into a central hub for your vision stack.

What you'll learn:

• Build Python plugins. Define plugin manifests and directory structures to register custom functionality within the FiftyOne ecosystem.
• Develop server side operators. Write functions to execute model inference, data cleaning, or metadata updates from the App interface.
• Build interactive panels. Create custom UI dashboards using to visualize model metrics or specialized dataset distributions.
• Manage operator execution contexts. Pass data between the App front end and your backend to build dynamic user workflows.
• Implement delegated execution. Configure background workers to handle long running data processing tasks without blocking the user interface.
• Build labeling integrations. Streamline the flow of data between FiftyOne and annotation platforms through custom triggers and ingestion scripts.
• Extend vector database support. Program custom connectors for external vector stores to enable semantic search across large sample datasets.
• Package and share plugins. Distribute your extensions internally and externally

Oct 08, 2026
9:00 AM - 11:00 AM PST
Online. Register for the Zoom!

Designing Multi‑Agent Systems: Sequential, Parallel, and Beyond with ADK

Multi‑agent systems are powerful but choosing the wrong interaction pattern can quickly lead to fragile, slow, or expensive AI systems.
In this talk, we explore the core multi‑agent design patterns enabled by ADK, including sequential, parallel, and more advanced coordination models. Rather than focusing on tools alone, we’ll look at how to think architecturally about agent collaboration.

You’ll learn:

• When sequential agents are the right choice and when they become a bottleneck
• How parallel agents improve speed and coverage (and the trade‑offs they introduce)
• Common failure modes in poorly designed agent interactions
• Practical criteria for choosing the right pattern based on task, latency, and reliability

By the end of the session, you’ll have a clear mental model for designing multi‑agent systems that are intentional, scalable, and production‑ready.

About the Speaker

Dr Roushanak Rahmat is an Enterprise AI Architect, Google Developer Expert (AI & Cloud), and recognized among the Top 100 Women in Tech (2025). With a PhD in Artificial Intelligence and over 15 years of experience, she specializes in designing and delivering enterprise-scale Generative AI, Agentic AI, and Deep Learning solutions that transform industries including healthcare, finance, energy, and public services.

Privacy by Deployment: Architecting Agent-Driven Localization Workflows for Regulated Environments

Most enterprise AI today is private by promise - a DPA, a SOC 2 report, or a contract clause that says, "we won't train on your data". For a regulated buyer, these are remedies after a breach, not controls that prevent or contain one. For organizations in healthcare, finance, defense, and government, privacy often requires stronger guarantees: data residency, customer-controlled execution, and, in some cases, operation within air-gapped environments.

This session demonstrates how agentic AI can automate a localization workflow while operating within these constraints. Using a real-world localization pipeline as an example, we will show how agentic systems can coordinate translation, review, quality assurance, and content preparation tasks while incorporating human checkpoints for approval and oversight.

We will also walk through the architectural patterns that enable these workflows to run inside customer-controlled and air-gapped environments without transferring sensitive content outside the customer boundary. The session includes a live product demonstration.

Key Takeaways

• Architectural patterns for deploying agentic AI in air-gapped and customer-controlled environments
• How agentic systems can automate localization workflows while preserving critical human review and approval processes
• Practical considerations for operating agentic workflows in regulated environments with auditability and governance requirements

About the Speaker

Shruti Joshi is building an AI powered secure localization stack for regulated industries such as healthcare, legal, finance that cannot send their content to a typical hosted SaaS. She brings 12+ years of engineering and architecture experience to the question this talk addresses: how do you make an agentic AI system deployable inside a regulated perimeter.

MCP Is the Interface; Skills Are the Operating Discipline

This talk shows how MCP and Agent Skills work together in practical agent systems. MCP gives agents a standard interface to tools, data, and workflows; skills encode the operating discipline that makes those connections reliable. Using a sanitized field-operations ledger as the case study, the talk walks through source intake, normalized state, uncertainty labels, role prompts, QA gates, and share-safe status drafting.

About the Speaker

Chuck Hernandez is an AI engineering and client-delivery leader with 10+ years across software, data platforms, and enterprise implementation, including 3+ years shipping production GenAI systems.

Agentic engineering is about good guidance.

Garbage Inn. Is garbage out? This is true. For many input and output processes. In biological life and in computer systems, and equally true when working with LLM’s. The better the prompt, the better the context, the better the focus, And the better the contextual awareness, the better the quality of the output the LLM’s generates.
This is the governance, art and practice of what we like to call agentic engineering, something I've been practicing over the last year.

About the Speaker

Dimitri Geelen builds things that don't need him once they're done. Frameworks, transitions, agentic systems — the measure of success is always the same: does it hold up when he leaves the room? He understands not just how to deploy, but what it takes for a new service to survive and scale inside a complex enterprise.

Oct 14, 2026
9:00 AM - 11:00 AM PST
Online. Register for the Zoom!

Agent as Policy for Robotic Manipulation

This talk will introduces how a general-purpose agent can directly drive a physical robot throughout task execution without any task-specific or environment-specific training. We introduce Agent as Policy (AGP), which places task planning and execution under the agent’s control.

About the Speaker

Xiaobai Liu is a Professor of Computer Science at San Diego State University (SDSU), where he directs the Machine Vision and Perception Lab. Prior to joining SDSU in 2015, he conducted research and taught at UCLA.

Grounding Multimodal Open-World Learning for Physical AI

Physical AI systems such as robots, unmanned vehicles, and embodied agents must perceive and reason about a world that is dynamic, unstructured, and rarely matches their training distribution. Yet most multimodal models remain brittle when confronted with novel objects, unseen conditions, and a messy unstructured environment.

This talk develops how grounding perception across modalities and environments can make open-world learning more robust for physically situated systems. Drawing on our recent work, I will highlight the core challenges of multimodal open-world learning including out-of-distribution detection, distribution shift generalization, and cascaded semantic grounding and point toward multimodal systems that stay reliable when deployed in the unpredictable open world.

About the Speakers

Salimeh Sekeh is an Associate Professor of Computer Science at San Diego State University (SDSU), where she directs the Sekeh Laboratory. Her recognition includes an NSF CAREER Award and a Cisco research gift (both 2022), and the Maine College of Engineering and Computing Early Career Research Award (2023), along with multiple industry and federal research awards.

Mary Wisell is a second-year PhD student in the Sekeh Lab and leads the lab's work on environment-aware OOD detection and cascaded failure analysis for multimodal intelligence with several publications in top-tier Machine Learning and computer vision conferences.