BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Meetup//Meetup Calendar 1.0//EN CALSCALE:GREGORIAN METHOD:PUBLISH NAME:Santa Cruz AI and Machine Learning Meetup Group X-WR-CALNAME:Santa Cruz AI and Machine Learning Meetup Group BEGIN:VTIMEZONE TZID:America/Los_Angeles TZURL:http://tzurl.org/zoneinfo-outlook/America/Los_Angeles X-LIC-LOCATION:America/Los_Angeles BEGIN:DAYLIGHT TZOFFSETFROM:-0800 TZOFFSETTO:-0700 TZNAME:PDT DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0700 TZOFFSETTO:-0800 TZNAME:PST DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT UID:event_316667091@meetup.com SEQUENCE:1 DTSTAMP:20261020T225051Z DTSTART;TZID=America/Los_Angeles:20261022T090000 DTEND;TZID=America/Los_Angeles:20261022T110000 SUMMARY:Oct 22 - Advances in AI at Virginia Tech DESCRIPTION:Santa Cruz AI and Machine Learning Meetup Group\nJoin our virt ual meetup to hear talks from AI researchers at Virginia Tech!\n\n**Date\, Time and Location**\n\nOct 22\, 2026\n9:00 AM - 11:00 AM PST\n**[Online. Register for the Zoom!](https://voxel51.com/events/advances-in-ai-at-virgi nia-tech-october-22-2026)**\n\n**Multi-Agent Communication: A framework\, diagnostic and mechanistic perspective**\n\nMulti-agent LLM systems are in creasingly used for collaborative reasoning\, debate\, and consensus\, yet their communication dynamics remain poorly understood. This talk presents a framework for studying multi-agent communication through diagnostic and mechanistic perspectives.\n\nI will discuss CONSENSAGENT\, which improves consensus by mitigating sycophancy\, alongside our diagnostic work on com munication patterns and failure modes in real-world multi-agent debates. I will then present ongoing work that moves toward a mechanistic understand ing of how these interaction patterns arise internally\, with the broader goal of making multi-agent systems more interpretable\, reliable\, and con trollable.\n\n*About the Speaker*\n\n[Priya Pitre](https://www.linkedin.co m/in/priya-p-a31360117/) I am an Ph.D student in the Computer Science Depa rtment at Virginia Tech (VT)\, co-advised by Dr. Xuan Wang and Dr. Naren R amakrishnan.\n\n**Exposing and Improving Fine-Grained Visual Grounding Abi lities of Lightweight Multimodal LLMs**\n\nLightweight multimodal LLMs can localize whole objects effectively\, yet often struggle when a query targ ets a small object part or fine-grained visual detail. This talk presents a reasoning-guided framework that teaches compact models to ground parts t hrough an explicit coarse-to-fine process: first locating the parent objec t\, then identifying the requested part.\n\nA part-aware reinforcement-lea rning objective provides stage-wise rewards for object accuracy\, part con tainment\, and the consistency of the model’s self-critique. Using these techniques\, a compact 4B-parameter model achieves state-of-the-art zero- shot part grounding while preserving its object-level performance.\n\nThes e advances can be used to enable lightweight MLLMs to support detail-orien ted tasks in biology and robotics.\n\n*About the Speaker*\n\n[Kazi Mehrab ](https://www.linkedin.com/in/ksmehrab/)I a CS PhD candidate at Virginia T ech\, where I currently focus on multimodal LLMs and computer vision tasks \, including visual perception\, reasoning and grounding.\n\n**Understandi ng Visual Generative Models for Precise Control**\n\nDespite remarkable pr ogress in image and video generation\, translating user intent into precis e and consistent visual outputs remains a challenge. This talk explores ho w understanding the representations within generative models can enable fi ner control over what they create.\n\nIt connects semantic image editing w ith compositional generation\, examining how visual concepts can be isolat ed\, manipulated\, and combined while preserving their identity and surrou nding content. Building on these insights\, structured visual inputs provi de a way to express complex intent through subject references\, poses\, an d spatial layouts.\n\nThe discussion then extends from images to video\, w here representations must evolve to preserve scene continuity while accomm odating motion and change. Together\, these directions establish a unified perspective on how visual representations can support controllable editin g\, composition\, and coherent generation across space and time.\n\n*About the Speaker*\n\n[Yusuf Dalva](https://www.linkedin.com/in/yusuf-dalva/) i s a Ph.D. candidate at Virginia Tech\, advised by Pinar Yanardag and affil iated with the Sanghani Center for Artificial Intelligence and Data Analyt ics. URL;VALUE=URI:https://www.meetup.com/santa-cruz-machine-learning-meetup-gr oup/events/316667091/ STATUS:CONFIRMED CREATED:20260922T180813Z LAST-MODIFIED:20260922T180813Z CLASS:PUBLIC END:VEVENT BEGIN:VEVENT UID:event_316758139@meetup.com SEQUENCE:1 DTSTAMP:20261020T225051Z DTSTART;TZID=America/Los_Angeles:20261029T090000 DTEND;TZID=America/Los_Angeles:20261029T110000 SUMMARY:Oct 29 - How to Build Vision Data Agents with Tools\, Skills\, and MCP DESCRIPTION:Santa Cruz AI and Machine Learning Meetup Group\nIn this sessi on\, you’ll learn how to build production-ready AI agents that can reaso n over your data\, automate complex tasks\, and integrate seamlessly into your existing stack using tools\, skills\, and the Model Context Protocol (MCP).\n\n**Date\, Time and Location**\n\nOct 29\, 2026\n9:00 AM - 10:00 A M PST\nOnline. **[Register for the Zoom!](https://voxel51.com/events/how-t o-build-vision-data-agents-with-tools-skills-and-mcp-october-29-2026)**\n\ nWe’ll walk through how modern agentic systems move beyond simple prompt s—leveraging structured tools like dataset operations\, embeddings\, eva luation pipelines\, and model execution to take real action. You’ll see how these agents can tag data\, run inference\, evaluate performance\, and surface insights automatically\, all within a unified workflow.\n\nBy com bining natural language interfaces with programmable building blocks\, tea ms can dramatically reduce manual effort\, accelerate experimentation\, an d unlock faster decision-making across the ML lifecycle.\n\nWhether you're building data-centric AI systems\, managing large-scale vision datasets\, or exploring agentic workflows for the first time\, this session will giv e you a practical blueprint for getting started. URL;VALUE=URI:https://www.meetup.com/santa-cruz-machine-learning-meetup-gr oup/events/316758139/ STATUS:CONFIRMED CREATED:20260928T231457Z LAST-MODIFIED:20260928T231457Z CLASS:PUBLIC END:VEVENT BEGIN:VEVENT UID:event_316768241@meetup.com SEQUENCE:1 DTSTAMP:20261020T225051Z DTSTART;TZID=America/Los_Angeles:20261105T090000 DTEND;TZID=America/Los_Angeles:20261105T110000 SUMMARY:Nov 5 - Visual AI Agriculture Meetup DESCRIPTION:Santa Cruz AI and Machine Learning Meetup Group\nJoin our virt ual meetup to hear talks from experts on cutting-edge topics at the inters ection of agriculture and AI.\n\n**Date\, Time and Location**\n\nNov 05\, 2026\n9:00 AM - 11:00 AM PST\nOnline. **[Register for the Zoom!](https://v oxel51.com/events/visual-ai-agriculture-meetup-november-5-2026)**\n\nTalks will include:\n\n**U-Net Framework for Micro-Scale Surface Damage Segment ation in High-Resolution Soybean Seed Imagery**\n\nAccurate detection of s urface-level seed damage is critical for soybean seed quality assurance\, yet automated micro-scale damage detection in high-resolution imagery rema ins an open challenge due to extreme spatial variability in damage scale\, severe class imbalance across damage types\, and the computational demand s of processing ultra-high resolution agricultural imagery at scale. This research addresses the semantic segmentation of fine-grained soybean seed surface defects\, such as wrinkles\, dark spots\, and general surface dama ge\, in 6048 × 4024 pixel images where target damages can be as small as 18 × 18 pixels\, using an augmented dataset of 77\,000 individual seed im ages.\n\nTo overcome the difficulties of micro-scale detection\, spatial s parsity\, and class confusion\, we propose a two-stage training and dual m odel inference framework built on an optimized U-Net architecture with a R esNet34 encoder. In the first stage\, a damage specialist model is trained using weighted loss functions and a class-balanced approach that prioriti zes damage classes.\n\nIn the second stage\, transfer learning is applied to initialize a healthy seed specialist model from Stage 1 weights\, with rebalanced class weights and tile probabilities that identify healthy seed s. During inference\, a confidence-gated damage filter suppresses low-conf idence predictions\, and healthy seed labels are assigned only when the sp ecialist model's confidence exceeds that of the damage model.\n\nThe two s pecialist models achieve validation accuracies of 94% and 98.53%\, respect ively\, and the combined inference system successfully detects and localiz es all three damage categories across unseen test images under conditions of extreme spatial variability and class imbalance. Qualitative evaluation confirms close alignment of predicted boundaries with ground truth annota tions across all damage types\, including dark spots.\n\nThese results dem onstrate that confidence-gated dual model inference can reduce class confu sion in imbalanced micro-scale segmentation\, advancing the feasibility of fine-grained automated seed quality inspection at the scale.\n\n*About th e Speaker*\n\n[Saurav Upadhyaya](https://www.linkedin.com/in/sauravupadhya ya/) is an AI/ML researcher with a Master's degree in Computer Science\, w hose work spans agricultural AI\, conversational systems\, and public heal th\, consistently translating advanced technology into tangible real-world impact.\n\n**From camera to drone: wildfire detection and integration at the edge**\n\nWildfires are getting larger and more expensive every year. But caught early\, a fire is a small job for a small team\; caught late\, nothing stops it. The cameras to catch them are already installed\, so the work is not necessary more sensors but making the ones we have act.\n\nTh is talk walks the whole chain: a small vision model running on the camera that flags smoke\, an integration layer that scores the alert against terr ain\, weather and history to decide whether it matters\, and a drone dispa tched to confirm before anyone commits a crew. We will cover what has to r un at the edge and why\, how small the model can get before it stops seein g smoke\, and what it takes to trust a detection nobody has looked at yet. \n\n*About the Speaker*\n\n[Maxime Carriere](https://voxel51.com/events/ww w.linkedin.com/in/maxime-carriere-871756161) is co-founder of Kernwerk in Berlin\, where he works on compressing AI models to run on cheap embedded hardware.\n\n**5\,000 Flights to Answer One Question: Can a Drone Sample a s Well as a Human?**\n\nEnvironmental testing still begins with a person w alking onto a site with a shovel. We built a system that turns a plain-lan guage brief into a sampling campaign\, flies it\, and returns soil\, water and vegetation samples with GPS\, timestamps and chain of custody intact: 5\,000 flights\, roughly 9\,000 miles and 570 flight hours over a 185-acr e site in Roatan\, much of it under canopy where GPS degrades and clearanc es are tight.\n\nI will walk through the vision and planning stack\, inclu ding multispectral and thermal canopy sensing\, terrain and land-cover map ping\, and the constraint solver that clears a route against airspace and battery budget before anything leaves the ground. Then the harder problem: evidentiary comparability.\n\nA drone-collected result means nothing to a regulator unless it agrees with a hand-collected split under the same met hod\, so I will show the dual-collection protocol we designed to test exac tly that. I will close on why calibrated uncertainty and knowing when a mo del should refuse to answer matter more than raw accuracy once the output lands in a regulatory filing.\n\n*About the Speaker*\n\n[Rohan Talwadia](h ttps://www.linkedin.com/in/rohantalwadia/) is co-founder and CTO of LabGPT \, which automates environmental field sampling using drones and edge anal ysis\, with results returned through ISO/IEC 17025-accredited partner labo ratories. URL;VALUE=URI:https://www.meetup.com/santa-cruz-machine-learning-meetup-gr oup/events/316768241/ STATUS:CONFIRMED CREATED:20260929T183723Z LAST-MODIFIED:20260929T183723Z CLASS:PUBLIC END:VEVENT BEGIN:VEVENT UID:event_316572576@meetup.com SEQUENCE:1 DTSTAMP:20261020T225051Z DTSTART;TZID=America/Los_Angeles:20261111T090000 DTEND;TZID=America/Los_Angeles:20261111T110000 SUMMARY:Nov 11 - AI\, ML and Computer Vision Meetup DESCRIPTION:Santa Cruz AI and Machine Learning Meetup Group\nJoin our virt ual meetup to hear talks from experts on cutting-edge topics across AI\, M L\, and computer vision.\n\n**Time\, Date and Location**\n\nNov 11\, 2026\ n9:00 AM - 11:00 AM PST\nOnline. **[Register for the Zoom!](https://voxel5 1.com/events/ai-ml-and-computer-vision-meetup-november-11-2026)**\n\n**Age ntic RAG: Beyond Retrieve-and-Generate**\n\nRetrieval-Augmented Generation (RAG) has become the default pattern for grounding large language models in external knowledge\, but most implementations still follow a rigid retr ieve-once\, generate-once pipeline — one that struggles with multi-hop q uestions\, ambiguous queries\, and knowing when its own retrieved context is insufficient. This talk introduces Agentic RAG\, where retrieval is tre ated as an action within an agent's reasoning loop rather than a fixed ups tream step.\n\nWe'll examine query decomposition and routing for breaking complex questions into targeted sub-queries\, self-reflective and correcti ve retrieval loops that let an agent judge and re-query its own results\, and tool-orchestration patterns (via MCP) that let retrieval sit alongside other agent actions like database lookups and API calls. Using a live arc hitecture — evolving a standard RAG chatbot into an agentic\, MCP-connec ted system — we'll walk through what changes in design\, and where these systems introduce new failure modes: grounding drift\, latency and cost f rom repeated retrieval loops\, and cases where a simpler RAG pipeline stil l wins.\n\nAttendees will leave with a practical framework for deciding wh en the added complexity of agentic RAG is worth it\, and a set of design p atterns for building it correctly.\n\n*About the Speaker*\n\n[Balaji Venka tasubramaniyar](https://www.linkedin.com/in/balaji-venkatasubramaniyar/) i s a Technical Lead at Wisdom Infotech\, leading a 15+ person engineering t eam delivering enterprise solutions. With 13+ years of experience in enter prise software and insurance technology\, he specializes in agentic AI sys tems\, RAG architectures\, and vector databases.\n\n**GeoAI for the Physic al World: Earth Observation\, Foundation Models\, and Urban Digital Twins* *\n\nEarth observation provides a unique form of computer vision for under standing the physical world at city to continental scales. In this talk\, I will show how satellite imagery\, geospatial data\, machine learning\, a nd foundation-model representations can be combined to characterize urban environments and environmental conditions.\n\nI will present UrbanScope Op en\, an open GeoAI digital-twin prototype integrating Earth observation wi th 3D buildings\, vegetation\, land-surface temperature\, air quality\, no ise\, population\, and other urban data. I will also share lessons from my research using geospatial foundation-model embeddings for environmental p rediction across Europe.\n\nThe talk will discuss how these approaches can contribute to increasingly multimodal AI systems capable of reasoning abo ut real-world environments.\n\n*About the Speaker*\n\n[Cesar Alvarez](http s://voxel51.com/events/cesar.alvarez@uni-a.de) is a researcher at the Univ ersity of Augsburg working at the intersection of GeoAI\, Earth observatio n\, remote sensing\, and environmental intelligence. His research applies machine learning\, computer vision\, and geospatial foundation models to p roblems including urban environments\, climate risk\, air quality\, and ag riculture.\n\n**Can agents get curious?**\n\nMost AI agents are good at an swering a question once we tell them exactly what to look for. The harder problem is building agents that can explore a complex dataset autonomously : generating hypotheses\, deciding which analyses are worth running\, allo cating additional compute when evidence is ambiguous\, and knowing when th ey have enough evidence to stop.\n\nIn this talk\, I’ll show an architec ture for autonomous research agents that combines structured knowledge\, i terative tool use\, and explicit evidence tracking to turn open-ended ques tions into a sequence of testable investigations. I’ll discuss practical lessons from building and evaluating these systems\, including why more t est-time compute does not automatically produce better research and how pr ovenance and evaluation can make long-running agents more reliable.\n\nI ’ll close with a live example of an agent exploring a dataset\, revising its hypotheses\, and choosing what to investigate next.\n\n*About the Spe aker*\n\n[Srivatsa P](https://www.linkedin.com/in/srivatsa-p) is a member of technical staff at Sigma Computing\, where he works on AI agents that r eason over complex enterprise data. Previously\, he worked on machine lear ning at Apple and conducted research at the MIT Media Lab\; outside of tra ditional ML\, he has also worked on mapping coral reefs through underwater imaging\, which sparked an enduring interest in how intelligent systems m ake sense of messy real-world data.\n\n**Does Cross-Validation Work in Dee p Learning?**\n\nCross-validation is a standard in hyperparameter optimiza tion in classical ML\, but is generally considered computationally prohibi tive and unnecessary in deep learning. We investigated whether CV produces 1. better DL models 2. better estimates of future performance of said mod els. We focused on medical image classification as an important deep learn ing application domain.\n\nOur results show that cross-validation consiste ntly and materially reduces the gap between development and independent te sting error for relatively small training datasets (up to about 1000 image s). The test performance improvement was generally modest. We discuss the implications for deep-learning practitioners\, including whether the incre ased computational cost of cross-validation is justified.\n\n*About the Sp eaker*\n\n[Ljubomir Buturovic](https://www.linkedin.com/in/ljubomir-buturo vic-798156/) is an AI scientist and biotech executive with extensive exper ience in clinical diagnostics\, including work at Inflammatix and Pathwork Diagnostics. His current research focuses on deep-learning image classifi cation\, infectious disease adjudication using LLMs\, and development of A I models for helping critically ill patients. URL;VALUE=URI:https://www.meetup.com/santa-cruz-machine-learning-meetup-gr oup/events/316572576/ STATUS:CONFIRMED CREATED:20260915T220412Z LAST-MODIFIED:20260915T220412Z CLASS:PUBLIC END:VEVENT BEGIN:VEVENT UID:event_316617427@meetup.com SEQUENCE:1 DTSTAMP:20261020T225051Z DTSTART;TZID=America/Los_Angeles:20261119T090000 DTEND;TZID=America/Los_Angeles:20261119T110000 SUMMARY:Nov 19 - Visualizing and Debugging Hugging Face's LeRobot Data DESCRIPTION:Santa Cruz AI and Machine Learning Meetup Group\nThis workshop is a hands-on tour of what actually breaks when you load real LeRobot dat asets\, and how Voxel51's native LeRobot support in FiftyOne catches it.\n \n**Date\, Time and Location**\n\nNov 19\, 2026\n9:00 AM - 10:00 AM PST\nO nline. **[Register for the Zoom!](https://voxel51.com/events/lerobot-works hop-november-19-2026)**\n\nRobot learning data doesn't look like the data that trains most vision-language models. It's temporal\, low-dimensional i n the parts that matter most\, embodiment-specific\, and expensive to coll ect\, and increasingly\, it's shipped in Hugging Face's LeRobot format. Bu t packaging robot data consistently isn't the same as being able to see it : episodes hide broken metadata\, silently dropped sensor modalities\, and mismatched schemas between robots that look identical on paper.\n\nStarti ng from the core differences that make robot data hard (temporal structure \, proprioception\, action grounding\, embodiment heterogeneity\, data sca rcity)\, we'll move into live debugging: a metadata bug that broke multipl e dataset imports\, sensor streams that vanish without an error\, and two robots running an identical benchmark whose data is structurally incompati ble. By the end\, you'll know what to check before you trust a LeRobot dat aset\, and how to actually look inside one. URL;VALUE=URI:https://www.meetup.com/santa-cruz-machine-learning-meetup-gr oup/events/316617427/ STATUS:CONFIRMED CREATED:20260918T224856Z LAST-MODIFIED:20260918T224856Z CLASS:PUBLIC END:VEVENT BEGIN:VEVENT UID:event_316757887@meetup.com SEQUENCE:1 DTSTAMP:20261020T225051Z DTSTART;TZID=America/Los_Angeles:20261210T090000 DTEND;TZID=America/Los_Angeles:20261210T110000 SUMMARY:Dec 10 - AI\, ML and Computer Vision Meetup DESCRIPTION:Santa Cruz AI and Machine Learning Meetup Group\nJoin our virt ual meetup to hear talks from experts on cutting-edge topics across AI\, M L\, and computer vision.\n\n**Date\, Time and Location**\n\nDec 10\, 2026\ n9:00 AM - 11:00 AM PST\nOnline. **[Register for the Zoom!](https://voxel5 1.com/events/ai-ml-and-computer-vision-meetup-december-10-2026)**\n\n**You r mAP is fine. Your slices are not. Finding the images that actually matte r**\n\nA single mAP can look acceptable while one slice is unusable\, smal l objects\, occlusion\, or a confused class. I will run a pretrained detec tor on a few hundred public images (COCO-style\, not a production crawl)\, report the headline metric\, then break the same run by slice and walk th rough the worst samples.\n\nThe claim is modest: you do not need a million images or a trained-from-scratch model to see what the mean is hiding\, a nd you cannot relabel everything\, so the next labeling budget should foll ow those slices. No employer data and no confidential dashboards.\n\n*Abou t the Speaker*\n\n[Amandeep Jiddewar](https://www.linkedin.com/in/amanfj/) is a Machine Learning Engineer at Pinterest\, with prior work in U.S. man ufacturing analytics (CertainTeed / Saint-Gobain) and marketplace allocati on (QuinStreet). Emory Goizueta MSBA (2019)\, Kaggle Expert. This talk is in his personal capacity and does not represent his current or prior emplo yers.\n\n**Production AI Agents: Tool Use\, Evaluation\, Guardrails and Ob servability**\n\nBuilding an AI agent is easy. Knowing whether it is behav ing reliably in production is much harder. This session explores how to ev aluate agents that use tools and external systems\, including tool-call va lidation\, tracing\, observability\, failure analysis\, guardrails\, permi ssions\, and human review.\n\nAttendees will learn practical patterns for understanding not just what an agent answered\, but why it acted and wheth er the action was safe.\n\n*About the Speaker*\n\n[Akshay Talathi](https:/ /www.linkedin.com/in/akshaytalathi/) is a Vice President at Goldman Sachs with over 10 years of experience building large-scale\, cloud-native\, and distributed systems across financial services and technology. His current focus is GenAI and agentic AI\, including production AI agents\, MCP\, en terprise integrations\, guardrails\, observability\, and reliable AI-drive n workflows.\n\n**From Virtual Worlds to Physical AI: Building Simulation Pipelines That Survive the Real World**\n\nRobots are moving beyond percep tion systems that only detect and classify objects toward systems that mus t understand a scene\, learn behaviors\, and act in the physical world. In this talk\, I’ll show a practical Physical AI workflow using OpenUSD\, BowerBot\, NVIDIA Isaac Sim and Isaac Lab to build a robot’s virtual wor ld\, including its body\, cameras\, sensors\, and environment.\n\nWe’ll explore how demonstrations and reinforcement learning can teach behaviors in simulation\, how vision and sensor data become observations for a robot policy\, and what needs to match when transferring that behavior to real hardware. The demo follows a small robot from a simulated environment towa rd a real-world task\, exposing both the power and the limitations of sim- to-real.\n\nThe goal is to make the path from pixels to actions concrete f or computer vision and machine learning practitioners.\n\n*About the Speak er*\n\n[Arturo Morales Rangel](https://www.linkedin.com/in/arturo-morales- rangel) is the founder of Binary Core LLC\, a Physical AI engineering comp any working across simulation\, perception\, robot learning\, and deployme nt.\n\n**Prompting Video-Language Models in Production: Lessons from Enter prise Video Tagging and Evaluation**\n\nVideo-language models can now desc ribe\, tag and answer questions about long-form video\, but production res ults depend heavily on prompt design and evaluation. This talk shares fiel d lessons from deploying such models for media\, news\, streaming and spor ts organizations\, including how to structure prompts for consistent\, sch ema-bound outputs across hours of footage.\n\nOn one sports video-tagging project\, consolidating the prompt definitions cut estimated processing co st about 3x\, and about 90% of the originally reported issues were later r esolved through prompting alone. I'll also walk through building evaluatio n frameworks for video outputs\, including for hard domains like egocentri c video\, and what changes when migrating between model versions.\n\nAtten dees will leave with reusable prompt patterns and an evaluation checklist they can apply to their own video pipelines.\n\n*About the Speaker*\n\n[Ya sh Gugale](https://www.linkedin.com/in/yashgugale/) is a Senior Forward De ployed Engineer at TwelveLabs in New York\, integrating video foundation m odels (Marengo for embeddings and search\, Pegasus for video-language unde rstanding) for media\, news\, streaming and sports customers. He previousl y spent about six years in video streaming engineering at Peloton and Sams ung TV Plus\, and holds an MS in Computer Graphics Technology from Purdue. URL;VALUE=URI:https://www.meetup.com/santa-cruz-machine-learning-meetup-gr oup/events/316757887/ STATUS:CONFIRMED CREATED:20260928T230354Z LAST-MODIFIED:20260928T230354Z CLASS:PUBLIC END:VEVENT X-ORIGINAL-URL:https://www.meetup.com/santa-cruz-machine-learning-meetup-g roup/events/ical/ X-WR-CALNAME:Santa Cruz AI and Machine Learning Meetup Group END:VCALENDAR