BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Palaner//palaner Single Event//EN
METHOD:PUBLISH
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:meetup-7331741ced@palaner.app
DTSTAMP:20260802T161017Z
DTSTART:20260804T220000Z
DTEND:20260804T233000Z
SUMMARY:Aug: Agentic on K8s + swag
LOCATION:90 Broadway
DESCRIPTION:Saved in Palaner as: saved\n\nPalaner match: 26% · why: lots of
  people going\n\nKomodor is sponsoring food & bev this time - thank you Ko
 modor!!\n\nAs usual:\n\n1. RSVP's close 48 hrs before the event. Please ma
 ke it easy on us by RSVP'ing only if you intend to show up.\n2. Don't be l
 ate! You'll get locked out\, and that's no fun!\n===\n\nAgenda\n\n6:00 pm:
  food / drinks / networking\n6:20 pm: Talk 1: Building Agentic AI with Mic
 kael Alliel\n7:00 pm: Who is hiring? Who is looking?\n7:10 pm: Talk 2: TBD
 \n7:50 pm: Talk 3: Build an AI agent\, win a prize with Harout Parseghian\
 n8:30 pm: Finish\n\nSpeakers & topics\n\n1. How to Build Quality-Driven Ag
 entic AI in Noisy Big Data Environments\, Mickael Alliel\, Komodor\n\nBuil
 ding reliable agentic AI systems in production environments presents uniqu
 e challenges when dealing with massive\, noisy datasets. This talk shares 
 hard-won lessons from developing Klaudia\, Komodor's AI agent that process
 es millions of Kubernetes events daily to deliver autonomous troubleshooti
 ng with 95%+ accuracy.\n\nThe fundamental challenge isn't LLM capability—i
 t's building systems that maintain reliability when 90% of your data is no
 ise. We'll explore why most agentic AI fails in production: hallucinations
  masquerading as insights\, inability to validate reasoning chains\, and t
 he brittle nature of RAG systems when dealing with complex\, interconnecte
 d failure modes.\n\nThis session covers practical know-how learned through
  painful production iterations: how to build validation frameworks that ca
 tch LLM errors before they reach users\, architectural patterns for constr
 aining problem spaces without losing effectiveness\, and techniques for cr
 eating evidence-based reasoning that can be audited and improved systemati
 cally.\n\nYou'll learn specific strategies for LLM validation in high-stak
 es environments\, including confidence scoring systems\, multi-agent verif
 ication patterns\, and iterative investigation loops that prevent runaway 
 reasoning. We'll cover the hard-earned lessons about what works and what s
 pectacularly fails when building trustworthy AI agents that must deliver a
 ccurate results rather than plausible-sounding explanations.\n\n2. TBD\\\,
  TBD\n\n3. Build an AI agent\\\, win a prize with Harout Parseghian\n\nHar
 out was 1 of 2 winners of our first AI Agent Build Contest. He has created
  an agent that scans container images and then rationalizes the result\, m
 aking a decision about whether to admit the images or not. Check it out! h
 ttps://github.com/haroutp/k8s-supply-chain-agent\n\nHarout can share a lot
  about his journey and his use of AI\, including:\n- the architecture of t
 he agent\n- the model calls made\n- the tool calls made by the agent\n- ev
 erything he learned building this\n- everything that broke along the way!\
 n\nAbout our sponsors\n\nkomodor\n\nHybrid/Remote option\n\nSorry\, this e
 vent is in-person only\, so please join us in Cambridge!
URL:https://www.meetup.com/boston-kubernetes-meetup/events/313866563/
CATEGORIES:palaner,saved
STATUS:TENTATIVE
END:VEVENT
END:VCALENDAR