devopsdays · ghoneycutt · Feb 28, 2026
@@ -0,0 +1,21 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "Git Up, Git Out: AI Answering the Call for Our Communities in Crisis"
+Type = "talk"
+Speakers = ["allison-higgins"]
++++
+
+In early November 2025, millions of Americans lost access to their SNAP
+benefits overnight due to a government shutdown, including thousands of
+families across Atlanta. In the face of this crisis, I decided to take action.
+Using Claude, React, and SCSS, I built an open source, live web application
+that lists food resources by county and zip code for Atlanta families in need.
+This talk walks through how I leveraged Claude, and a small but growing list of
+resources to take swift action for the community to design, architect, and
+maintain a small, but growing civic tech platform. I will discuss the project’s
+technical architecture, my Claude workflows, my collaborators, and the lessons
+I learned about using AI responsibly for swift action for the community. Most
+importantly, I want to encourage other engineers and researchers to start their
+own civic technical projects as a way for expertise to meet real-world needs.
@@ -0,0 +1,25 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "Streamlining AI Native Product Evals: LLM as Judge, Human in the Loop, and Reliable Agent Testing"
+Type = "talk"
+Speakers = ["anil-pantangi"]
++++
+
+Enterprises deploying AI components need more than metrics. They need reliable
+systems that developers can trust and iterate on quickly. In this talk, I will
+share how to design product evaluation pipelines tailored for AI native
+development workflows. We will explore how to utilize LLM as a judgment
+technique, combined with human-in-the-loop feedback, to assess and refine
+generated outputs in a manner that is both repeatable and efficient for
+engineering teams.
+
+The session will cover how to transform evaluation into a developer-first
+workflow, including patterns for parallel evaluation runs, reliability
+tracking, and regression alerts. We will also discuss how to maintain
+compliance and audibility in industries where trust and oversight are crucial.
+
+As a forward look, I will share how these evaluation practices build the
+foundation for agentic systems that improve themselves and how they prepare
+teams for future advancements and demanding customer needs.
@@ -0,0 +1,19 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "404 Community Not Found? Let’s Fix That!"
+Type = "talk"
+Speakers = ["calypso-hernandez"]
++++
+
+Success in tech isn’t just about what you know—it’s also about who you know and
+how you engage with the world around you. In this session, we’ll explore why
+building a strong community is essential for personal and professional growth.
+From expanding your network to sharpening your skills, being part of a tech
+community can open doors, spark collaborations, and provide the motivation
+needed to push through challenges. We’ll discuss the power of reciprocity—how
+staying engaged and giving back can lead to unexpected opportunities, new
+career paths, and meaningful connections. Whether you're looking to grow,
+pivot, or find support during tough times, investing in your community is an
+investment in yourself.
@@ -0,0 +1,15 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "Building and using communities to further your AI career"
+Type = "talk"
+Speakers = ["danny-castro"]
++++
+
+In this session I will discuss how communities such as AI tinkerers or various
+meetup groups will help lead to your career growing as it is easy to be
+isolated. Working in a silo is a thing but that won't make you a better
+developer and it won't make you a better AI professional compared to those who
+have feedback loops, can bounce ideas off people. So I intend to show them how
+to get of their shell and become a part of these communities.
@@ -0,0 +1,16 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "Code Reading Club"
+Type = "talk"
+Speakers = ["elizabeth-jamison"]
++++
+
+Inspired by and using Felienne Herman's Code Reading Club template, last year
+we started a Code Reading Club here in Atlanta. Every month we look at an
+unfamiliar piece of code (in a perhaps unfamiliar language) and, as a group,
+practice our code reading skills, an increasingly important skill in this age
+of AI generated code. In this talk I'll explain how Code Reading Club works,
+what benefits our participants have found, and how you can start your own Code
+Reading Club at work, in your community, or with some friends.
@@ -0,0 +1,16 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "When AI Hits Production Everything Changes"
+Type = "talk"
+Speakers = ["hajira-sultana"]
++++
+
+AI systems often behave well in demos but fail in surprising ways once they
+reach production. In this Ignite talk, I will share real lessons from deploying
+AI systems and why many of them break under real users, real data, and real
+pressure. I will highlight common issues like missing observability,
+over-trusting AI outputs, and ignoring human behavior in system design. This
+talk is about what happens after launch and what DevOps teams can do to reduce
+risk, improve resilience, and design AI systems that work in the real world.
@@ -0,0 +1,19 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "Running Tenstorrent in DevOps"
+Type = "talk"
+Speakers = ["jared-rhodes"]
+
++++
+
+AI is changing our delivery pipelines, but most teams default to GPUs without
+evaluating alternatives. In this Ignite, I’ll show what it actually takes to
+pilot Tenstorrent in a DevOps environment: where Wormhole PCIe cards fit, how
+the TTNN/PyTorch path and TT-Forge compiler plug into CI, and what SREs should
+watch (power, thermals, scheduling, and failure modes). We’ll cover
+observability hooks, image build basics, rollback plans, and how to stage a
+low-risk bake-off alongside your current stack. No hype—just a clear checklist
+to help platform, SRE, and MLOps teams decide if and how Tenstorrent belongs in
+their fleet.
@@ -0,0 +1,17 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "The Continuous Integration of Self: 8 Japanese Habits for Sustainable Engineering"
+Type = "talk"
+Speakers = ["jason-torres"]
++++
+
+In tech, we value the power of the compound effect. Small commits that lead to
+massive releases. Yet, when it comes to our own lives, we often ignore this
+logic, opting for radical overhauls that lead to burnout. This session
+introduces eight ancient Japanese habits—including Kaizen (micro-improvements)
+and Hara Hachi Bu (the 80% rule)—reimagined as an "Operating System" for the
+modern developer. You will learn how to "refactor" your daily routine to
+improve focus, increase career longevity, and replace the exhaustion of the
+"hustle" with a system of intentional, incremental growth.
@@ -0,0 +1,28 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "The Vulniverse: Now with AI! (Hold the Slop Please)"
+Type = "talk"
+Speakers = ["jeff-diecks"]
++++
+
+Finding, fixing, and sharing software vulnerabilities is a challenge on the
+easiest of days. With all of the standards, formats, channels, and
+personalities, it can feel like an insurmountable hill to climb. Just when we
+thought we had sorted it out with the Vulniverse Alphabet Soup Guide to help
+clear up the Coordinated Vulnerability Disclosure (CVD) process, along comes
+new adventures thanks to AI in the form of Cyber Reasoning Systems (CRS). These
+new tools are capable of autonomously finding and fixing bugs in open source
+software. How will CRS’s impact the daily work of Product Security & Incident
+Response Teams (PSIRTs), Security Researchers, Computer Emergency Response
+Teams (CERTs), and Corporate Incident Response & Security Teams (CSIRTs)? How
+will all of this be received (or rejected) by open source project maintainers?
+What can we do to keep humans involved and prevent AI slop?
+
+This session will explore the new capabilities and real world results from the
+open source CRS’s created through DARPA and ARPA-H’s AI Cyber Challenge
+(AIxCC). With the experience from dozens of zero-days being found and reported
+responsibly (with generated patches also being provided), we’ll review how the
+process and tools have been received (or rejected) by a cross-section of open
+source projects, and what it means for the future of CVD.
@@ -0,0 +1,25 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "The Future DevEx is Autonomous"
+Type = "talk"
+Speakers = ["kiah-imani"]
++++
+
+AI agents are no longer science fiction. They can already test, debug, and even
+deploy applications with minimal human intervention. This shift means developer
+experience (DX) is evolving beyond human-first design.
+
+Today, DX focuses on people through onboarding docs, tutorials, and SDKs.
+Tomorrow, it must also focus on AI through APIs, schemas, and metadata that
+agents can interpret, reason about, and act upon without manual guidance.
+
+OpenAPI, JSON Schema, and GraphQL introspection are not just conveniences for
+developers. They are machine-readable blueprints that AI agents can use to
+understand and interact with your system.
+
+In this talk, we will explore what it means to build AI-first developer
+experiences, why your next developer audience may not be human, and how you can
+prepare your platform for a world where the best developer on your team is an
+agent.
@@ -0,0 +1,55 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "Storytelling as a Strategic Tool in Tech Leadership"
+Type = "talk"
+Speakers = ["kimasia-ayers"]
++++
+
+In today’s technology-driven world, innovation alone is not enough. The ability
+to inspire, align, and mobilize teams, stakeholders, and entire ecosystems
+often hinges on one thing: the story you tell. Storytelling is no longer a soft
+skill, it is a strategic leadership tool that shapes perception, influences
+decision-making, and drives adoption across products, processes, and projects.
+Leaders who harness the power of narrative can translate complex ideas into
+shared vision, turn data into action, and build cultures that thrive on
+clarity, trust, and purpose
+
+This session explores how storytelling functions at the intersection of human
+behavior, organizational dynamics, and technological innovation. Participants
+will engage with frameworks such as the Axis of Assimilation, which uncovers
+how social expectations shape organizational behavior; Spiderweb Theory, which
+maps influence, power, and information flow within networks; and Social Capital
+Theory, which highlights how relationships amplify messages and drive
+organizational outcomes. Through real-world examples from product launches,
+cross-functional projects, and global initiatives, attendees will learn how to
+craft narratives that resonate with diverse audiences, bridge silos, and
+accelerate impact.
+
+Beyond communication, storytelling becomes a lever for leadership: it shapes
+culture, establishes credibility, and drives collective action. Participants
+will leave with actionable strategies to articulate vision, influence decisions
+at all levels, and embed narrative into leadership practices that guide product
+strategy, team alignment, and stakeholder engagement. By mastering
+storytelling, leaders can transform abstract concepts into actionable insight,
+foster inclusive environments, and inspire innovation that resonates across
+markets and communities.
+
+Key Takeaways
+1. Narrative as Influence: Understand how strategic storytelling guides
+   decision-making, aligns teams, and drives adoption of initiatives and
+   innovations.
+2. Map and Leverage Networks: Apply Spiderweb Theory and Social Capital Theory
+   to identify key stakeholders, influence information flow, and amplify
+   impact.
+3. Lead Through Insight: Use the Axis of Assimilation to craft narratives that
+   navigate social and organizational dynamics, building trust, clarity, and
+   cohesion.
+
+By the end of this session, participants will be equipped to elevate
+storytelling from communication tool to leadership instrument. They will leave
+ready to influence projects, products, and processes, align diverse teams
+around vision and strategy, and embed narrative into the core of tech
+leadership, creating environments where innovation, engagement, and impact
+flourish.
@@ -0,0 +1,18 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "The Shallow End of Deep Learning"
+Type = "talk"
+Speakers = ["mandi-walls"]
++++
+
+Massive LLMs take huge amounts of data to train and datacenters to run. You
+have some trend data from your services, and maybe it would be kinda
+interesting to have some way to do some predictions or projections off that
+data. How far could you get with some Python and a Jupyter notebook?
+
+Far enough to be dangerous in your next production performance meeting. In this
+talk, we'll wade into the shallow end of the deep learning tools ecosystem and
+talk through data practices for getting your own insights out of small AI
+projects.
@@ -0,0 +1,28 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "Stop Fighting Your AI: Engineering Prompts That Actually Work"
+Type = "talk"
+Speakers = ["martin-rojas"]
++++
+
+Your prompts suck. Your AI outputs prove it.
+
+Most developers treat prompt engineering like magic - throw words at the AI and
+hope for the best. But getting consistent, production-quality results from AI
+requires actual engineering principles, not creative writing. This session
+reveals the systematic approach to prompt engineering that produces reliable
+results: using AI itself to debug your prompts, building constraint systems
+that prevent garbage outputs, and creating prompt architectures that scale
+across your entire development workflow.
+
+You'll master:
+- Why "be creative" prompts fail (and the structural patterns that don't)
+- How to reverse-engineer successful prompts using AI feedback loops
+- Constraint engineering that eliminates 90% of useless AI outputs
+- The prompt versioning and testing strategies that actually work in production
+
+Perfect for: Developers integrating AI into workflows, technical writers
+automating documentation, and anyone tired of playing prompt lottery with
+inconsistent results.
@@ -0,0 +1,28 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "The Day Claude Code Deleted My Cluster: A Cautionary Tale About AI Guardrails"
+Type = "talk"
+Speakers = ["michael-forrester"]
++++
+
+"You have full access to the pipeline. Do what you need to do."
+Famous last words.
+
+In this lightning talk, I'll share the hilarious (and horrifying) story of what
+happened when I gave Claude Code full pipeline access and stepped away for 30
+seconds. When I came back it had not only completely wrecked the Kubernetes
+cluster but not even two troubleshooting systems sessions later it wrecked
+almost every network card in the set of Linux systems.
+
+This is a story about nondeterministic systems, the illusion of AI
+understanding, and why "the AI knows what it's doing" is the most dangerous
+phrase in modern DevOps. I'll share the actual troubleshooting spiral that
+escalated from "let me help" to "I've destroyed your cluster and systems," and
+the guardrails I now enforce religiously.
+
+5 minutes. 20 slides. And one very hilarious probably very blameful post-mortem
+with Claude Code afterwards. One very expensive lesson about trusting AI agents
+with infrastructure access even if it was just for a short while. Come for the
+disaster. Stay for the wisdom.
@@ -0,0 +1,15 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "Your Outdated Docs are Costly: Why You should be writing tests for your docs"
+Type = "talk"
+Speakers = ["omari-gaskins-jr"]
++++
+
+As a software engineer for 5 years at JPMC, I’ve had to onboard to hundreds of
+apps, and one consistent issue I always ran into, was incorrect, or completely
+unrelated readmes and/or onboarding docs. After talking to several fellow
+engineers, it seems no one has tried to offer a feasible solution. This seemed
+weird to me, as this is a universally accepted problem, and that’s why I’ve
+created the solution, and would love to share it with you all.
@@ -0,0 +1,10 @@
++++
+Talk_date = ""
+Talk_start_time = ""
+Talk_end_time = ""
+Title = "My Experience Homelabbing"
+Type = "talk"
+Speakers = ["paul-vinueza"]
++++
+
+I started my homelab journey back In December and would like to share my experience.