Accepting submissions
Not accepting submissions
Submission guidance
When submitting, please tell us:
- What problem your session addresses
- Who the session is for
- What the audience will learn
- Whether the work is in production, experimental, or in progress
- What technical details, demo, case study, or story you will share
- What makes the session useful to practitioners
Please also add tags to help us understand your proposal. Tags may describe domains, techniques, systems, or formats, for example:
#observability #inference #agents #security #governance #finops #sre #devops #platformengineering #mlops #databases #scalability #kubernetes #developerplatforms #failurestory #demo #workshop #casestudy #workinprogress
|
SS
Swapnil Prakash Sankla [WIP] One Dataset, Three Physical Representations (YOUR SCHEMA decides how much work the query engine does at runtime)Session title One Dataset, Three Physical Representations more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
AS
Aditya Singh When the Client Disconnects but the Agent Keeps RunningRootconf 2026 Proposal Draft Session title When the Client Disconnects but the Agent Keeps Running more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
VK
Vivek Kalyanarangan Decoding at a million tokens when the KV cache lives in host memory: what actually bounds the stepDecoding at a million tokens when the KV cache lives in host memory: what actually bounds the step more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
SM
Shaurya Madukuri We spent a week optimising an LLM pipeline and most of what we measured was noiseOne-line summary A 12,858-image vision pipeline took 1h51m, so we tested three optimisations against it. Our first measurements said none of them worked, because run-to-run variance was 43%. Once we controlled the noise, two were real speedups (up to 7x from URI image transport) and one traded latency for cost. Here’s how we told them apart, and why the rest of the wins came from reading code rat… more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
KANIKA SINGHAL When Agents Go Rogue: Building a Zero-Trust Agent MeshWhen Agents Go Rogue: Building a Zero-Trust Agent Mesh more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
MS
Mourjo Sen Test Cases Push Humans Out of the Loop, Test Properties InsteadAbstract AI generates code faster than we can even read it. If we cannot read the code fast enough, we are unlikely to catch any of the quality gaps in its code or tests. Essentially, relying on the speed of LLMs reduces our role to a passive overseer - the irnoy is that our cognitive abilities do not catch logical bugs as just a passive overseer [1]. We must find a way to become active problem s… more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
Pronomita Dey (WIP) A Platform Story: Designing a Delete Path That Survives Being WrongSummary Determining which of 470,000 repositories are still alive is a harder classification problem than it looks, and getting it wrong takes down production. How we model liveness as a graph property, why read activity is the signal that matters, and how we redesigned deletion to be survivable. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
GM
Gaurav Maheshwari Surviving the query-pocalypse!Proposal template We have a new user persona - AI Agents. What does it mean for platform availability, reliability & performance? more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
OA
Om Asnani Agents Without Judgment Rituals: Failure Modes When Teams Adopt AI Into Learning and Platform OpsPlatform and learning teams are shipping agents into real workflows. Curriculum helpers, support triage, docs bots, onboarding copilots. The failure mode I keep seeing is not that the model is dumb. It is adoption without judgment rituals. No clear verify step. No escalation path when the agent is confidently wrong. No review of near-misses. Training that rewards speed of output over quality of d… more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
sm
sarthak makhija Transpiling the Untranspileable: Patterns for Building Clean TranspilersTranspiling the Untranspileable: Patterns for Building Clean Transpilers more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
PG
Prakhar Goel Streaming Through Failure: Designing Reliable Transcription StreamsReconnecting a WebSocket restores the connection, but not the state of an interrupted stream. At Adalat AI, we faced this in live courtroom transcription; I’ll show how we retransmit audio the server missed and replay transcript updates the client missed after a network drop. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
M
Mahendra [Add a catchy title or a Work-in-Progress (WIP) title]How Thousands of Offline-First AI Agents Can Learn Together more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
Alosh Denny Your Model Does Not Need Floating Point, But It Does Need One Thing You Probably SkippedOne-line summary. Open weight models quantise to eight bit fixed point after training, with no retraining and no calibration set, and keep their tool calling intact, provided you get one detail right that silently destroys the model if you miss it. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
Alosh Denny It Scored 100 Percent Because It Had Stopped AnsweringOne-line summary. A quantised model scored a perfect 100 on one category of a function calling benchmark while emitting tool calls on 1.2 percent of prompts, and the headline average made it look merely mediocre. This is a talk about validating a model change before it reaches production. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
Alosh Denny Break A Model, Then Fix It: A Hands-on Session On Quantisation And Proving It WorkedOne-line summary. Participants quantise a small open weight model to eight bits, watch it break in the specific way that looks like a precision problem but is not, fix it, and then build the evaluation that would have caught the breakage. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
Alosh Denny Who Owns The Precision Decision, And How Does Your Team Prove It Was Safe?One-line summary. Quantisation changes the model, but the decision usually sits with whoever owns the serving cost. I want to compare notes on who actually makes that call across teams, and what evidence they require before it ships. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
SN
Suraj Nath TraceQL Internals: Building a Query Language for Distributed TracesTitle: TraceQL Internals: Building a Query Language for Distributed Traces more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
SN
Suraj Nath Metrics from Traces at Scale: Surviving the Cardinality ExplosionSession title Metrics from Traces at Scale: Surviving the Cardinality Explosion more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
VA
Vaibhav Arora Stop prompting. Start harnessing: how I run AI agents that don’t fall overAbstract Most demos show an agent completing one happy path. Real work is messier: long tasks, bad tool calls, context that overflows, and “it worked yesterday” failures. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
P
Prasun AI On-Call Engineer for Enterprises - Architecture OverviewTopic: AI On-Call Engineer for Enterprises Description: As AI speeds up SDLC, more code gets shipped with less engineers to manage it in prod. How do we ensure the bottleneck does not just shift to prod? How can we leverage AI to create a stable production environment and resolve incidents faster in the new age aoftware delivery era? more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
Anant Shrivastava Owning Your AI: Building a Private Stack You Can Actually Live WithOne-line summary This session documents the journey of buiding and operating AI / LLM systems on hardware you directly control be it home or office. The key focus is on model choices, trust, power, isolation, maintainability and real-world usefulness. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
VK
Vikash Kumar When ML Serving Becomes a Distributed Systems ProblemDeploying an ML model is easy. Operating hundreds of them reliably is not. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
UK
Utkarsh Kanwat Your Agent Can Run as Long as You Can Check ItSession title Your Agent Can Run as Long as You Can Check It more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
saurabh hirani Observing AI apps: know your instrumentationSummary As we helped customers instrument their AI applications, the tooling matured alongside the work, and each engagement pushed us to cover more surface area and get more visibility: from vanilla OpenTelemetry on RAG apps, to OpenLLMetry, the Bifrost AI gateway, and OpenLIT for agents, and now Langfuse. We rebuilt each stage on an open-source reference setup to show what each approach sees, w… more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
PR
Prudhvi Raj The GPU Was Waiting on the CPU: Taking the Host Off the LLM Serving PathSession title The GPU Was Waiting on the CPU: Taking the Host Off the LLM Serving Path more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
Ayushman Bhattacharya What Should an AI Search System Remember, Reuse, and Recompute?Session title What Should an AI Search System Remember, Reuse, and Recompute? more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
MC
Manas Chaturvedi From AI Agents to AI Platform: Lessons From Building an In-House Multi-Agent PlatformFrom AI Agents to AI Platform: Lessons From Building an In-House Multi-Agent Platform more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
PM
Pankaj Merisha Bare-Metal Agents: Building a Low-Latency Inference Stack for In-House Coding LLMsBare-Metal Agents: Building a Low-Latency Inference Stack for In-House Coding LLMs more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
HK
Hemant Kumar Benchmarking AI SRE agents before they touch productionSummary Over the past few months we’ve been evaluating AI SRE agents with several of our customers: OpenSRE, HolmesGPT, K8sGPT and a handful of others. We started with the things any SRE would check before trusting one of these on a real incident. That only took us so far, so we built a benchmark on top of the open-source SREGym framework, where the agent’s diagnosis is graded against what actual… more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
KS
Kushagar Sharma Hermes in Production: Building an AI Infrastructure Engineer That We Could Actually TrustOne-line summary How we built and productionized Hermes for a customer in the AI-powered industrial operations and fleet intelligence space: an infrastructure agent that investigates real systems, reasons about changes, and creates production-ready GitOps PRs while keeping execution deterministic, auditable, and safe. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
KJ
Karan Jagtiani Your Platform's Newest Tenant Is a Coding Agent You Didn't WriteYour Platform’s Newest Tenant Is a Coding Agent You Didn’t Write more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
SS
Shubham Shrivastav  Can an AI agent say "don't ship"? Building a release gate you can trustSession title An AI agent as a release gate: a ship or don’t-ship call you can trust more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
MK
Md Tarique Yaseen Khan Agents Are Easy, Memory Is HardAgents Are Easy, Memory Is Hard: What Our SRE Agent Is Allowed to Remember, and How We Check It more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
s
simran Breaking the ClassifierBreaking the Classifier: Hands-On Adversarial Evasion Against Machine Learning Systems more
I am submitting: To teach a workshop
I have a submission for: Hands-on workshop - 2-4 hours
|
|
SB
Sayan Bhattacharyya Your on-call agent has amnesia: building AI that learns from production incidentsSession title Your on-call agent has amnesia: building AI that learns from production incidents more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
h
heetgala Stop Shipping AI on Vibes: AI Evals for AgentsAbout this session Every team building with LLMs has lived the same story. The demo works beautifully, everyone’s impressed, it ships. Then a real user asks a slightly different question and the answer comes back confidently wrong. more
I am submitting: To show a demo
I have a submission for: A talk - 30-40 mins
|
|
HJ
Harsh Joshi [Add a catchy title or a Work-in-Progress (WIP) title]Session title Your AI Agent Is a Distributed System more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
GR
Guda Sai Vikhyath Reddy Engineering Beyond the Broker: Solving Operational Challenges Without Replacing KafkaProposal template Session title Kafka: The Infrastructure Around the Infrastructure more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
Sreeram Venkitesh OTel Memory Management 101: Lessons learnt while running OpenTelemetry scraping 20k clusters in productionOTel Memory Management 101: Lessons learnt while running OpenTelemetry scraping 20k clusters in production more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
SP
Silu Panda The Authentication Boundary Before an AI Gateway StartsOne-line summary A public-code case study of a LiteLLM Redis-cluster startup failure: why synchronous cluster construction needed credentials before a later IAM connection callback could supply them, and how to test the real authentication handshake. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
RA
Rajanarayana A Silent Cluster Decay: Building an In-Cluster Operator for Autonomous MaintenanceOne line summary A Kubernetes cluster that is Ready is not a cluster that stays healthy. etcd fragments, nodes drift, registries go stale, and a central control plane is the wrong place to fix any of that. This session covers CLuster Maintainer (CLM): an in-cluster operator that runs periodic maintenance and reconciles drift so a multi-node cluster stays stable without humans. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
J
Jitendra Upgrade Broke My Query… Break Free with AI [ Detect, Debug, Fix with AI-Assisted Self-Service Platform ]Session title Upgrade Broke My Query… Break Free with AI Detect, Debug, Fix with AI-Assisted Self-Service Platform more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
JM
Justin Mclean What happens when a node diesProposal template You do not need to answer everything perfectly in the first pass. Start with what you know. Short, rough answers are fine. The editorial team may follow up to help shape the proposal further. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
SM
Sai Khalandar Mokula Why LLM citations break across multiple retrievers: the one architectural rule to fix attribution and unlock precise groundedness evalsSession title Why LLM citations break across multiple retrievers: the one architectural rule to fix attribution and unlock precise groundedness evals more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
MS
Manish Sharma "Mise en Place" for Kubernetes: Why We Stopped Fetching Images at RuntimeOne-line summary Your cluster isn’t failing to pull an image; your release failed to pack it. This talk shows how we turned a fragmented image scavenger hunt into one verifiable bundle for predictable Kubernetes deployments, upgrades even in airgap environments. more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
sk
sandeep kella [Add a catchy title or a Work-in-Progress (WIP) title]Proposal template You do not need to answer everything perfectly in the first pass. Start with what you know. Short, rough answers are fine. The editorial team may follow up to help shape the proposal further. more
I am submitting: To speak
I have a submission for: Show a demo - 5-10 mins
|
|
G
Ganesh Building FlareDB, an Apache Beam native streaming database.Overview: FlareDB is an Apache Beam native streaming database for running Beam data pipelines. It’s built in Rust and it uses a streams-tables architecture inspired by the ideas described in the Streaming Systems book (Chapter 6). more
I am submitting: To show a demo
I have a submission for: A talk - 30-40 mins
|
|
AP
Adithya Parthasarthy 3 Weeks to 3 Hours: Building an AI Co-Pilot for Kubernetes Upgrade Preparation3 Weeks to 3 Hours: Building an AI Co-Pilot for Kubernetes Upgrade Preparation at Booking.com more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|
|
TN
Tushar Naik How PhonePe’s in-house Code Review platform (Arnold) helped control the Code FloodProposal template How PhonePe’s in-house Code Review platform (Arnold) helped control the Code Flood more
I am submitting: To speak
I have a submission for: A talk - 30-40 mins
|