~/conference

Open Source Summit India 2026

Open Source Summit India unites the open source ecosystem, offering a platform for collaboration, knowledge-sharing, and exploring innovative solutions.

titleauthorlength
Keynote: Linus Torvalds in Conversation with Dirk HohndelLinus Torvalds, Dirk Hohndel36m
Linux 7.1 ships as steady incremental progress on a strict 9–10 week release cadence, with the merge window handled as roughly 200 merges in two weeks and a…
Interview at Jasmine Hall 2 (Level 3) on
Guide To Become Linux Kernel MaintainerKrzysztof Kozlowski38m
Linux kernel maintainership requires being an active contributor, knowing Git and the email-based workflow, and committing real time to reviews and grooming…
Talk at Lotus Ballroom 3 (Level 3) on
Trust, but Sandbox: Securing AI Agents in KubernetesSudhanshu Prajapati, Prateek Mishra39m
Shows an exploit chain where a curl-installed agent reads a .env, posts secrets to a pastebin and deletes a workspace via a planner→worker A2A delegation…
Talk at Jasmine Hall 2 (Level 3) on
Hey Yocto, Build Me a Custom Embedded Linux! Er, NoKaiwan Billimoria42m
Shows how Yocto plus BitBake actually builds a full custom embedded Linux: downloads ~8 GB, compiles for ~3–4 hours using a powerful build host and eats ~77…
Talk at 204 (Level 2) on
How Kubernetes Networking Really Works: A Packet’s Journey Across Pods and NodesAshwin Sriram, M Viswanath Sai51m
Live demos build a kernel‑level mental model: a hand‑rolled tun + user‑space encapsulator works but loses ~25% of packets and adds 43 bytes per packet plus CPU…
Talk at 204 (Level 2) on
Open Source Is Not the Same AnymoreFaeka Ansari, Hrittik Roy40m
Open source projects now change licenses, lock APIs, or stay controlled by one vendor; curl ends a $86,000 bug‑bounty after AI‑generated noise; Synadia tries…
Talk at 204 (Level 2) on
Kubernetes OIDC That Works in Practice: Keycloak + RBAC + Kubelogin Without Day‑2 PainManik Bindlish31m
Configures Kubernetes API with OIDC args (client_id, issuer URL, CA cert) and uses Keycloak + RBAC + kubelogin so logins return 5-minute tokens that kubelogin…
Talk at 205 (Level 2) on
Running On-Device AI With Qualcomm AI Engine Using LiteRT & ExecuTorchKartikey Rawat39m
Claims ability to run full PyTorch and TFLite models completely on-device using ExecuTorch + Qualcomm QNN (via LiteRT), with an AOT compile → partition →…
Talk at Jasmine Hall 2 (Level 3) on
Keynote Panel: Open Source Innovation Powering India’s Digital Public InfrastructureRadha K, Suresh M K, Julian Gordon30m
Panel describes a deployed, production trust stack: DigiYatra runs on 38 airports, 24 million users, 100 million uses and replaces a central repository with no…
Panel at Jasmine Hall 2 (Level 3) on
I Break Things, AI Fixes Them: Building a Self-Healing CI/CD PipelinePremved Dhote34m
Claims a working proof-of-concept called Pipeline Doctor that uses GitHub Actions to trigger a LangGraph orchestration with LangGraph + Gemini for analysis and…
Talk at Lotus Ballroom 3 (Level 3) on
Security: Why It _has_to Be Open SourceMike Bursell40m
Argument: security software works better open because independent review, reproducible signing and supply-chain tooling scale trust—examples include Sigstore…
Talk at 204 (Level 2) on
Recipes and Runtimes: Making Sense of Containers in 2026Soundarya Rangarajan55m
Builds a simple Node + Express app and shows how a naive Ubuntu-based Docker image balloons to ~1.79 GB, then drops to ~330 MB with a multi-stage Node 24 slim…
Talk at 204 (Level 2) on
AI as a Platform Engineer: Explaining Kubernetes Failures, Not Just Detecting ThemParanitharan Kalaiselvan38m
Comcast describes an AI explanation layer that cuts Kubernetes deployment troubleshooting from 30–45 minutes to about 5 by combining declarative JSON rules…
Talk at Lotus Ballroom 3 (Level 3) on
CI/CD, APIs, and Scaling: What Every Cloud Native Developer Needs To Kn...Aditya Soni, Aditi Gupta39m
Shows a concrete pipeline: code -> Tekton triggers -> Kaniko/Buildah builds -> Trivy scans -> images pushed to Quay/Harbor and deployed by Argo CD as GitOps so…
Talk at 204 (Level 2) on
Keep AI on Track: Guardrails + OpenTelemetry ObservabilityPrabal Rakshit40m
Shows a concrete pipeline that instruments guardrails across UI, orchestration, LLM, retrieval and infra using OpenTelemetry and an LGTM backend…
Talk at Jasmine Hall 2 (Level 3) on
Operating a Self-Healing Bare-Metal Kubernetes Platform at Global ScaleAparna Prabhu, Nikhil Pathak40m
DigitalOcean describes a self-healing bare-metal Kubernetes platform running storage workloads across 15 regions, 366 nodes, 5,000+ pods with hardware from 40…
Talk at 205 (Level 2) on
Lightning Talk: If Zephyr Wants To Power AI Cameras, What Must Change?Rutvij Trivedi18m
Zephyr’s video API lacks per-frame metadata and graceful stream shutdown, which causes NPU inference to target the wrong display frame (observed as N+2 frame…
Lightning at Lotus Ballroom 2 (Level 3) on
Lightning Talk: Strengthening Zephyr’s Camera Framework: Architecture Review and EnhancementsElgin Perumbilly, Ankit Siddhapura12m
Adds runtime PM to Zephyr’s camera stack by implementing PM action callbacks in the IMX335 driver and testing on an STM32N6 board, so the sensor’s clocks and…
Lightning at Lotus Ballroom 2 (Level 3) on
DTS 101: From Roots To Trees, Aka Devicetree for BeginnersKrzysztof Kozlowski40m
Step‑by‑step practical guide to writing DTS and bindings for Linux: explains exact YAML layout (filename must match compatible), how to express a…
Talk at Lotus Ballroom 2 (Level 3) on
Scientific Machine Learning With NVIDIA PhysicsNeMoAniket Kulkarni, Samudyata Minasandra31m
Shows a workflow that converts OpenFOAM simulations into a trained neural operator that predicts surface pressure, temperature, and velocity in seconds. Uses…
Talk at Jasmine Hall 2 (Level 3) on
Downstream Zephyr RTOS Release Management - Keeping up With Upstream PaceParthiban N43m
Zephyr reaches ten years with ~3,000 contributors, ~150k commits and ~15k commits per release, and downstream teams face fast upstream churn: 17,000 PRs in one…
Talk at Lotus Ballroom 2 (Level 3) on
An Introduction To Coreboot and LinuxBoot: Building Modern Open Boot StackManish Baing, Arun Mahendran49m
coreboot initializes CPU, DRAM and PCI up to a usable RAM stage and then hands control to LinuxBoot which embeds a tiny Linux kernel plus initramfs and uses…
Talk at 206 (Level 2) on
Sponsored Session: Does Zephyr Scare the Bare Metal Embedded Developer World?Khasim Syed Mohammed, Soumya Tripathy47m
Zephyr claims to structure complexity rather than hide it and maps bare‑metal idioms to equivalents so familiar patterns still work. Hands‑on evidence: simple…
Talk at 206 (Level 2) on
Building Hybrid Quantum-Classical Pipelines : A Practic...Sainath S, Guncha M, Divya S, Amutamil E42m
Hybrid pipeline that strings classical HPC and quantum hardware together using Slurm + Prefect orchestration and a QRMI middleware to route jobs to QPUs via…
Talk on
Decoding the Open-source Blueprint for India's Sovereign AI FutureVincent Caldeira, Abhishek Kumar Singh41m
A production-ready sovereign stack called M4 coding runs on Indian infrastructure using OpenStack + OpenShift, Kubernetes, multi-GPU nodes (eight GPUs per…
Talk at 205 (Level 2) on
Building a Zero-Copy DSP Offload Framework in Linux Using RPM...Vishnu Pratap Singh, Paresh Bhagat37m
A practical upstream-friendly framework uses remoteproc, rpmsg and ALSA with DMA-backed shared memory and a zero-copy ping-pong buffer design to avoid…
Talk at Lotus Ballroom 2 (Level 3) on
Event-Driven Platform Engineering: From Reactive Ops To Autonomous Control LoopsJosephine Eskaline Joyce, Prashanth Bhat48m
A reference architecture shows Prometheus alerts converted to CloudEvents, routed through Kafka/NATS/Knative, evaluated by OPA/Kyverno for policy and then…
Talk at 206 (Level 2) on
Hardening IoMT Medical Devices: Defense‑in‑Depth on Yocto‑Based Embedded LinuxAbraham Gogulamudi48m
Yocto-based hardening for IoMT lays out a defense‑in‑depth build: hardware root of trust and secure boot with TPM/OTP fuses, minimal U‑Boot and kernel configs…
Talk at Lotus Ballroom 2 (Level 3) on
Judgement Day: Benchmarking "Black Box" LLMs With Open Legal DatasetsKannan Murugapandian40m
A controlled experiment evaluates GPT-5, Gemini 2.5 Pro and DeepSeek‑R‑One on LegalBench using a custom async Python harness that runs 4,302 reasoning traces…
Talk at Jasmine Hall 2 (Level 3) on
Keynote: The Open Source Runway: Building the Foundation for the Agentic EraToddy Mladenov10m
Argues that production-grade autonomous agents need a five-layer runway—Linux, Kubernetes, agent runtime, orchestration, harness—with concrete components…
Talk at Jasmine Hall 2 (Level 3) on
Bridging the Gap To Autonomous AI-Native TelcoGanesh Narayan, Praveen Kumar Kalapatapu47m
Claim that 95% of telcos use AI but 79% remain below autonomy level 2 and need a leap to closed‑loop autonomy. Presents a concrete blueprint: a common agent…
Talk at 206 (Level 2) on
Efficient Performance Profiling for Virtual MachinesSandipan Das43m
Mediated PMU in KVM (merged in Linux 7.0 mediated PMU) gives guests direct access to hardware PMU registers with mutual exclusion so profiling accuracy matches…
Talk at Lotus Ballroom 3 (Level 3) on
Panel: India’s Leading Role in Developing and Deploying Open Source Technologies Powering Digital Trust InfrastructureAnkita P, Kamlesh N, Smita S, Rajesh D, Denver D46m
India shows concrete deployments: a telecom DPI cut spam for 600 million subscribers; DigiYatra runs reusable KYC on blockchain; Bhutan ships Credible identity…
Panel at 205 (Level 2) on
How To Tame Your Agents?Devidas Jadhav40m
Chronological account of building agentic pipelines from copy‑paste prompts to a multiagent factory, claiming a 16‑agent system built a C compiler and showing…
Talk at Jasmine Hall 2 (Level 3) on
Performance-by-Design: Embedding Intelligent Scaling and Guardrails Into Platform EngineeringJosephine Eskaline Joyce, Tanya Shanker45m
Claims a concrete method: codify SLOs, instrument app-aware metrics with Prometheus/OpenTelemetry, and enforce policies via OPA so autoscalers (HPA or KEDA)…
Talk at Lotus Ballroom 3 (Level 3) on
Hardware Assisted PMU VirtualizationManali Shukla26m
Hardware-assisted PMU virtualization extends the upstream Mediated PMU framework with automatic save/restore of guest PMU state in hardware, direct interrupt…
Talk at Lotus Ballroom 3 (Level 3) on
The Maintainer’s Perfect Storm: Survival, Succession, and a Decoupled FutureAmrit Kumar Verma, Gaurav Mishra25m
SW360 is mid-migration from a Liferay monolith to Next.js migration, which halts releases and exposes a decade of tribal knowledge loss when several veteran…
Talk at 206 (Level 2) on
Keynote: From Consumption to Global Leadership - Open Source SummitArpit Joshipura23m
India is number two in open source developer contributions and the fastest-growing community at ~100% year-over-year, tracked via LFX Insights that shows…
Talk at Jasmine Hall 2 (Level 3) on
Learnings From a $1M Discovery and Funding ExperimentAnsh Arora30m
floss.fund launches late 2024 as a $1M annual experiment to finance critical FOSS projects and introduces funding.json as an open, crawlable manifest…
Talk at 205 (Level 2) on
Strengthening India's FOSS CommunityAnsh Arora39m
FOSS United registers in 2020 and runs three programs for individuals, communities, organizations, backing ~20 city chapters and 50+ student groups while…
Talk at 205 (Level 2) on
Beyond Static Benchmarks: Chaos Based Agent...Saramma G, Suganya S, Chetana A, Praveen K, Deepak S40m
Claims an open source AgentCert (ACE) framework that runs Litmus Chaos-driven fault injections, collects telemetry with LangFuse, and certifies agents by…
Talk on
Keynote: Rust and Linux​: How the Rust ​Language is ​Going to ​Help Linux ​Succeed in the ​FutureGreg Kroah-Hartman18m
Rust code already reaches 113,000 lines in the kernel mostly as C bindings, and new drivers (notably graphics and Android Binder) are being accepted in Rust…
Talk at Jasmine Hall 2 (Level 3) on
Demystifying PCI Interrupts: Understanding MSI/MSI‑X in LinuxShradha Gupta42m
Explains why legacy INTx fails at scale and shows MSI (up to 32 vectors) and MSI‑X with 2048 vectors that let each queue have its own interrupt, and how Linux…
Talk at 206 (Level 2) on
Pull Once, Scale Everywhere: Fixing Image Pull Bottlenecks With DragonflyShivani Rathod29m
Dragonfly replaces thousands of concurrent registry pulls with a cluster-local peer-to-peer distribution using manager, scheduler, seed peer and per-node…
Talk at 206 (Level 2) on
Towards a Tool for Access-Affinity Based Structure Reordering in the Linux KernelMadadi Vineeth Reddy, Aboorva Devarajan40m
A prototype tool traces field-level loads/stores on struct rq, builds an access-affinity graph by counting temporal co‑access within a short sliding window…
Talk at Lotus Ballroom 3 (Level 3) on
Syzbot To Mainline: How I Merged 21 Kernel Patches as a First-Time ContributorDeepanshu Kartikey21m
21 patches across 15+ kernel subsystems (ext4, gfs2, btrfs, ocfs2, f2fs, mm, tracing, networking, BPF) merge to mainline using syzbot reports as seeds. Fixes…
Talk at Jasmine Hall 2 (Level 3) on
“Hey AI, Train Llama”: Making Kubeflow Agent-Native With MCPAkash Jaiswal, Abhijeet Dhumal41m
Kubeflow MCP Server exposes the Kubeflow SDK as MCP-callable tools so agents can submit and manage training via natural language while enforcing cluster…
Talk at Lotus Ballroom 3 (Level 3) on
53 Years of Ethernet: Evolving With Open Standards for AI InfrastructureKapil Mehta41m
History and limits of Ethernet lead to three concrete standards efforts: UET 1.0 (11 Jun 2025) proposes a transport replacing TCP/UDP to provide in-protocol…
Talk at 205 (Level 2) on
Beyond the First PR: Why Contributors Stay in Open Source CommunitiesYashvant Singh14m
Retention depends on belonging not complexity: communities that give responsive feedback on PRs, open DMs on Slack, weekly meetings and mentorship keep…
Talk at 204 (Level 2) on
From Noise To Signal: Building Smart Observability Pipelines With OpenTelemetryManoj Sardana, Saloni Narang, Saiyam Pathak45m
Live demo builds an OpenTelemetry pipeline that enriches Kubelet metrics with deployment and image labels, filters out readiness/health-check traces with OTTL…
Talk at 206 (Level 2) on
Democratizing Grid Intelligence for Developing Nations Through Linux-open-source InfrastructureSatabdy Jena, Nikita Rao38m
An open Linux/FOSS blueprint named Sonio core assembles modular services—network topology processor, probabilistic load forecast, online load flow, state…
Talk at Jasmine Hall 2 (Level 3) on
Don’t Trash It, Hack It: Reverse Engineering Secrets & Re-purposing ISP RoutersDheeraj Reddy Jonnalagadda41m
Shows how a commercial ISP router is fully reclaimed: firmware is dumped via UART or SPI (CH341), U‑Boot is intercepted because of no U‑Boot timeout, the boot…
Talk at Lotus Ballroom 2 (Level 3) on
A Practical Perfetto Introduction for AOSP and Linux DevelopersStefan Lengfeld41m
A hands‑on Perfetto primer that shows how to record traces on Android and embedded Linux using a Yocto tutorial, captures kernel ftrace events and custom…
Talk at Lotus Ballroom 2 (Level 3) on
Panel Discussion: A Balancing Act: OSPOs in Action Within BFSI (Banking, Financial Services, and Insurance)Ram I, Srividya G, Madhusudanan GK, Charudutta Panchmatia28m
Banks describe how a Log4j crisis forces asset discovery and rapid mitigation, so OSPOs inventory components with SBOMs mandated by RBI SEBI CERT-IN, run daily…
Panel at 205 (Level 2) on
Keynote: Accelerating Innovation Through Open SourceAayush Bhatnagar16m
Describes how an operator builds production-grade cloud native stacks: contribution to ONAP, running India’s largest 5G standalone network, and a developer org…
Talk at Jasmine Hall 2 (Level 3) on
JSON Wastes 60% of Your AI-LLM Tokens. TOON Fixes ThatVitthal Mirji25m
JSON adds structural noise that commonly consumes 40-60% of LLM tokens, for example a 100-row payload that became ~127,000 tokens as pretty JSON. TOON…
Talk at Jasmine Hall 2 (Level 3) on
Writing for Machines: How To Capture Your Project's "Vibe" and Survive AI SlopKaushlendra Pratap Singh, Shaheem Azmal M MD43m
Maintainers report that AI increases churn not velocity: an OSS project rises from ~120–150 PRs/month to 300–400, with merge rate collapsing from 85% to 7.5%…
Talk at Jasmine Hall 2 (Level 3) on
Generic BootLoader on Android PlatformsNaina Mehta33m
Generic BootLoader (GBL) is a Rust-based bootloader in AOSP built as a UEFI application that standardizes boot flow across ARM64 x86 and RISC‑V and plugs into…
Talk at Lotus Ballroom 2 (Level 3) on
The Containerization of the Operating System: Exploring Bootc and the Future of Linux DeploymentHema Arun44m
Claims that a full Linux host becomes an OCI image managed like an app and that updates/rollbacks are atomic and fast. Uses bootc with Podman (Podman Desktop +…
Talk at 206 (Level 2) on
Oops, My AI Agent Just Deleted All My Email: Locking Down Agents With Signed PoliciesRahul Vishwakarma31m
Aflock enforces cryptographic guardrails for agents by requiring signed policy files that declare allowed tools, files, APIs, and flow rules; policies emit…
Talk at 204 (Level 2) on
But It Builds on My Machine!: Building Deterministic OCI Images With NixBenjamin Philip40m
Explains how Nix turns image builds into deterministic artifacts by hashing full dependency closures, sandboxing builds (no network) and storing outputs in a…
Talk at 203 (Level 2) on
Breaking Valkey on Purpose: Chaos Fuzzing a High-Performance Key-Value Store With Agentic AIRenuka Uttarala40m
Demonstrates a custom Chaos Fuzzer that spins up Valkey clusters (example run: seven shards, five replicas per shard = 42 nodes), injects SIGKILL/SIGTERM and…
Talk at 203 (Level 2) on
Quantum-Safe TLS in Practice With Open Quantum Safe & OpenSSL 3Divyanshu Agrawal, Shubham Bhardwaj, Anitha Natarajan47m
Demo environment uses OpenSSL 3.5 on a Fedora 43 container and builds liboqs plus oqs-provider 0.10 to add candidate algorithms. A self‑signed ML‑DSA CA/leaf…
Talk at 205 (Level 2) on
Breaking the TCP Barrier: Accelerated I/O for S3 With RDMAVidushi Mishra24m
Validation uses SEF + Ceph RGW + RADOS behind NFS Ganesha with Mellanox RDMA NICs (25 Gbps); four clients (each 1 GB) show TCP 426 Mbps vs RDMA 428 Mbps on a…
Talk at Lotus Ballroom 2 (Level 3) on
Poisoning the Well: Why AI Governance Is the OSPO’s New FrontierMadhusudanan GK25m
Banks face three concrete attacks: a RAG pipeline invisible instruction that approves a $500,000 credit via hidden white‑on‑white PDF text, a Trojan‑horse…
Talk at 205 (Level 2) on
Full Disk Encryption for Confidential Computing GuestsAnirban Sinha45m
Argument: confidential VMs need disk confidentiality as well as memory protection; the proposed stack combines dm-verity + LUKS with systemd/dracut boot flow…
Talk at 206 (Level 2) on
Agentic Delivery, Guardrailed: AI in CI/CD & Platform Engineering (Open Source Patterns)Manas Ray, Kalyan Kolachala44m
A reference architecture that fuses Git + Jenkins/Tekton + Argo CD with OpenTelemetry (Tempo, Loki, Prometheus/Mimir, Grafana) and a delivery intelligence…
Talk at Lotus Ballroom 2 (Level 3) on
ESim – Democratizing Electronic Design Automation Through Open SourceSumanto Kar, Shanthi Priya30m
eSim is a Linux‑first open EDA stack that actually bundles KiCad + ngspice for analog, adds Makerchip integration for RTL, and supports Sky130 PDK and IHP for…
Talk at Lotus Ballroom 2 (Level 3) on
Secure by Default: Building an AI-Augmented, OSS-Powered Reusable CI/CD PipelineJenisten Xavier40m
Reusable GitHub Actions/GitLab CI pipeline that 150+ repositories adopt with zero extra config, enforcing GitLeaks secret scanning, OSV Scanner for new…
Talk at Lotus Ballroom 3 (Level 3) on
Pruning Kernel CVEs With Code Reachability AnalysisAshish Bijlani, Chandni B32m
Static reachability maps NVD CVEs to vulnerable kernel functions using Kconfig, module load status and interprocedural call graphs and the open‑source KCV…
Talk at 203 (Level 2) on
Meet FRED: The Future Face of X86 Event ProcessingShivansh Dhiman23m
Claims a hardware redesign that replaces the IDT with a single MSR entry and a uniform stack frame that carries an event data field and vector/type tags, adds…
Talk at Lotus Ballroom 3 (Level 3) on
Detecting and Defusing the Ticking Time Bomb: Understanding CVEs and Upstreaming in YoctoSiddharth Doshi35m
Claims a semi‑automated scanner that merges CVE feeds from three to five sources into a common database, compares that daily to an SBOM and recipe list, and…
Talk at Lotus Ballroom 2 (Level 3) on
Keynote: Cut Database Costs, Fund Innovation with ValkeyRoberto Luna Rojas06m
Valkey is a BSD‑licensed, Linux Foundation‑stewarded in‑memory key‑value engine used as a look‑aside cache in front of MySQL, MariaDB and PostgreSQL…
Talk at Jasmine Hall 2 (Level 3) on
Facing Extinction: The Fight To Keep Hardware Crypto Engines in Embedded LinuxKamlesh Gurudasani, T Pratham30m
Upstream maintainers are removing kernel support for crypto engines after benchmarks show CPU crypto (ARMv8/AVX512) outperforms accelerators by 12–56×…
Talk at Lotus Ballroom 3 (Level 3) on
Confidently Wrong: When AI Cannot Catch Its Own BugsShailja Thakur44m
Models regularly produce simple logic bugs—arithmetic, timezones, unit conversions—and then verify their own wrong answers. An empirical study using Queen 30B…
Talk at Jasmine Hall 2 (Level 3) on
The Next Evolution of Java: ☕️ Achieving Hyper Performance and Efficiency in Cloud Native WorkloadsDaniel Oh53m
Compares concrete techniques: Quarkus + Jib for minimal images, GraalVM Native Image for tiny cold starts (example build ~25–30s with ~2–3GB peak build RAM…
Talk at 206 (Level 2) on
Introducing In-Kernel PSI Auto Monitor FeaturePintu Kumar Agarwal32m
Optional in-kernel PSI Auto Monitor captures thread-level contributors exactly when configured PSI thresholds breach, using existing PSI tracepoints and a…
Talk at Lotus Ballroom 3 (Level 3) on
Beyond SBOMs and SLSA: Build Trust With Custom Attestations for End-to-End Supply Chain SecurityLavakush Biyani, Pranay Shah21m
Supply-chain blind spots get specific fixes: supply chain attacks rise 742% since 2019, recent Trivy compromise hijacks 76 of 77 tags and renames 44 repos, and…
Talk at 204 (Level 2) on
Keynote: Welcome BackArpit Joshipura04m
Shows a concrete setup: the Linux Foundation database feeding custom AI build agents that run analytics, surface surprising metrics about contribution growth…
Talk at Jasmine Hall 2 (Level 3) on
Speeding up Your ML Workload: Pytorch Compile and Distributed TrainingAishwariya C, Priyanka N, Kavya G, G Chander34m
Explains how torch.compile turns eager PyTorch into graph mode using torchdynamo, AOT Autograd and torch inductor and what each stage actually emits and…
Talk at 204 (Level 2) on
Identity and Access Management for the Decade AheadThivaharan Kalyanasundaram35m
Modern IAM must treat AI agents as first-class entities with unique credentials, lifecycle and auditable delegation; Thunder ID demonstrates this with a…
Talk at 205 (Level 2) on
From Pipelines To Provenance: Reproducible Builds With TektonShubham Bhardwaj, Divyanshu Agrawal33m
Two independent Tekton pipeline runs first produce different container SHAs, then after concrete fixes produce the same SHA (E59): pin the base image by…
Talk at 204 (Level 2) on
Governing the Ungovernable: Security and Compliance for AI Agents in Open Source Projec...Ronit Raj33m
Describes a working open-source governance layer called GitMesh that intercepts agent actions, compiles YAML policies to OPA, and records immutable…
Talk at 205 (Level 2) on
The Process of Exploration in AI Research: A Researcher’s PerspectiveBibekananda Hati30m
Describes Pile of Flow, an open‑source local experiment manager that converts code blocks into component-as-dictionary entries, centralizes configs, logs and…
Talk at Jasmine Hall 2 (Level 3) on
Keynote: Open By Design - Open Ecosystem, Governed AI, Trusted OutcomesGeeta Gurnani16m
IBM marks 115 years and cites concrete open projects: Lightwell for AI security (IBM + Red Hat + community), an open AI Risk Atlas tool that scores use-case…
Talk at Jasmine Hall 2 (Level 3) on
"I Didn't Peek: and I Can Prove It": Confidential Computing for Audits and RegulatorsMike Bursell39m
Confidential computing uses chip-level TEEs in Intel AMD Nvidia and Arm to keep data encrypted in use and to produce remote attestation signatures that…
Talk at 206 (Level 2) on
Meaning Over Bits: How Semantic AI and eBPF Define the 6G Open StackKhushi Chhillar, Satyam Soni14m
6G shifts from raw-bit throughput to goal-oriented semantic communication, embedding intent in-network with semantic encoder/decoder models running as native…
Talk at 205 (Level 2) on
Operators Live in the Past: Designing Reliable Kubernetes ControllersSomeshwaran Mohan Kumar46m
Operators read from a stale cache and that causes real race conditions at scale (examples: 600k containers, 20 controllers, 13 CRDs, clusters with 1,000…
Talk at 203 (Level 2) on
When Nobody Owns Quality: Making Testing Responsibility Explicit in Open Source ProjectsUjjwal Kumar Singh38m
Open source projects often have CI and tests but no clear enforcement, so quality stalls until governance changes. Evidence: Kafka drops from 90% failing main…
Talk at Lotus Ballroom 3 (Level 3) on
What If Npm Install Could Say No? Real-Time Defense Against Malicious PackagesSahil Bansal20m
Claim: a package manager can block malware at install time by running a threat-intel proxy for per-package verdicts, executing package scripts inside Landlock…
Talk at 204 (Level 2) on
Keynote: The Moat Moved from Code to ContextManish Dixit15m
Claims that AI makes writing code nearly free and that the moat moved from code to context backed by field data: Linux Foundation deploys an MCP server, LFX…
Talk at Jasmine Hall 2 (Level 3) on
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.