DevConf.IN 2026
2026-02-13..2026-02-14
Pune, India
| title | author | length |
|---|---|---|
| DevOps at LLM Speed: Using an AI Copilot for Kubernetes and Jenkins | Karan Jagtiani | 34m |
Claims a 50% cut in incident response time at Storylane by using Skyflo.ai, an open-source, self-hosted AI copilot that links Kubernetes, Jenkins, Argo, Helm… | ||
| Lost in Silence: Understanding When Screen Readers Don’t Speak Up | Ayushi Midha, Aishwarya Urne | 37m |
Demonstrates that missing announcements (notably on dynamic updates) break UX even when visual tests pass, affecting ~1.3 billion people (16% of the world)… | ||
| Bridging DevOps and MLOps: Unifying Pipelines with KitOps and GitOps | Neel Shah | 32m |
Claims that KitOps packages models code datasets and metadata into a versionable OCI artifact and that GitOps then gives automated, auditable deployments. Demo… | ||
| How to attack AI systems (and how to defend them) !!!! | Huzaifa Sidhpurwala | 42m |
Shows concrete attacks and concrete mitigations: data poisoning of public crawls that flips dog images into “fish,” physical adversarial patterns (printed dots… | ||
| MPI Meets Machine Learning: Unlocking PyTorch distributed for scaling AI workloads | Mansi Agarwal | 38m |
Maps MPI collectives onto PyTorch primitives and shows concrete tradeoffs: uses torch.distributed with backends NCCL Gloo or MPI, runs via torchrun/elastic… | ||
| Nomad: Lightweight Orchestration That Complements Kubernetes | Shaheen Sayyed | 12m |
Claims Nomad is a lightweight HashiCorp orchestrator delivered as a single binary that schedules Docker containers, virtual machines and raw executables, not… | ||
| Beyond ImagePullBackOff: A Stateless, Secret-less Distributed Registry for Edge | Prasanth Baskar | 23m |
Harbor Satellite presents a stateless, secret-less registry that pushes image material to edge nodes via a central Ground Control which emits OCI state and… | ||
| Battery Range Prediction using Federated Learning on Edge | Sagar Sundaray, Vinod Pathangay | 14m |
Uses a simulator to generate time series of uphill/downhill and auxiliary-load scenarios and trains a PyTorch predictive model locally on constrained ECUs… | ||
| The Cybersecurity Trifecta: Privacy, Governance, and Data Sovereignty in Action | Dhanashri Wategaonkar | 23m |
Claim: recent DPDP Act (Nov 2025) adds time‑limited, revocable consent so users can share data for “10 days” or “one day” instead of forever; presenter asserts… | ||
| Lightweight Observability with Performance Co-Pilot and Grafana | Ayushi Tiwari | 28m |
Shows a working observability pipeline that uses Performance Co-Pilot (pmcd + pmlogger) to read cgroup metrics (CPU, memory, disk I/O, network) and Grafana… | ||
| Talking Chaos: An AI Co-Pilot for Resilience With LitmusChaos | Pritesh Kiri | 36m |
An AI co‑pilot ties a code assistant to LitmusChaos so commands like “simulate a pod failure on payments” generate and run YAML experiments, return logs and a… | ||
| Shift-Left Security in Practice with Gitlab | Rashi Chaubal | 17m |
Shows a concrete CI/CD pipeline built on GitLab CI/CD that runs SAST, Semgrep, secret detection, dependency scanning and Trivy container scans (plus optional… | ||
| Reasoning Operators: Bringing LLM Logic Into Kubernetes Control Loops | Pratik Kumar Panda | 14m |
Demonstrates a safe pattern for adding an LLM reasoning layer to Kubernetes controllers: gather events, logs and CRD state, build a structured prompt, call… | ||
| Prompt Engineering 2.0 | Vedant Kabra | 15m |
Claims that Gemini and Claude outperform Cursor and other vibe coding platforms for longer context work, and recommends using divide and conquer and meta… | ||
| Design Systems at Scale: Maintaining UI Consistency in Dynamic Plugin Architecture | Mitesh Kumar | 19m |
Red Hat Developer Hub (RHDH) built on Backstage runs 100 plus plugins and demonstrates concrete answers: a centralized TypeScript theme package exporting… | ||
| Want to extend GitOps to Infrastructure and Workloads? Here's how. | Eeshaan Shreyas Sawant | 16m |
Most teams keep Kubernetes GitOps in Argo/Flux while infrastructure and serverless remain in brittle CI pipelines; PipeCD claims to close that gap by treating… | ||
| Cognitive A11y: Designing Neuroaccessible docs for Neurodivergent minds | Yash Guddeti, Kalyani Desai | 16m |
Documentation practice that cuts cognitive load and saves time: empirical claim that many readers skip about 80% of dense text to hunt a single command; four… | ||
| Building Secure & Reliable Edge Images Using RHEL for Edge Blueprinting & OSTree | Rajesh Dulhani, Dipayan Dutta | 33m |
RHEL for Edge uses Image Builder, blueprints, podman/containerfiles, composer CLI and OSTree to produce immutable QCOW2/VMDK images; the demo builds a v1 image… | ||
| Kubernetes for Machine Learning: Managing and Scaling AI Workloads | Drishti Jain | 42m |
Walks through a concrete MNIST pipeline built with Kubeflow: local Jupyter experiments, TensorFlow CNN training deployed as a TFJob CRD on Kubernetes, training… | ||
| Vaani:Architecting Conversational AI The end-2-end Pipeline of Modern Voice Agents | Abhishek Jha | 18m |
Vaani converts a text agent into a real‑time voice agent using noise suppression, VAD, whisper-medium for ASR and Piper for TTS, runs ASR/LLM workloads in… | ||
| Governing Global Vulnerabilities: Standards for Open Source Resilience | Yogesh Mittal | 37m |
Global vulnerability handling now pairs CVE identifiers with machine-readable ingredients and decisions: SBOM lists components, CSAF carries advisory metadata… | ||
| Anubis Will Judge You: AI-Powered Firewall Defending Web Traffic | Shane Cardoz M | 10m |
Anubis claims to replace CAPTCHAs by issuing a token after three checks: fingerprint verification, JS behavior challenge, and a proof-of-work CPU puzzle that… | ||
| Kubernetes Confessions: Real Mistakes We’ve All Made (and What They Teach Us) | Shashank Pai | 36m |
Seven reproducible failure modes with fixes: missing CPU/memory Vertical Pod Autoscaler recommendations and Prometheus history cause an analytic pod to grab 8… | ||
| Sustaining Yesterday’s Linux: Delivering Security for Today’s Workloads | Suyash Nalawade | 17m |
Long-term maintenance teams track upstream CVEs, assess applicability with a 0–10 severity score, and perform targeted backporting CVEs into decade-old… | ||
| Platform Wars: The Battle Between Golden Paths and Spaghetti Pipelines | Prithvi Raj Venkat Raj, Bharath N R | 40m |
Internal platforms often bloat with redundant tools — one cited enterprise buys 200 unused licenses — so the proposed fix is an opinionated, five-step… | ||
| Image Mode: The Supported Way to Extend RHEL CoreOS | Prachiti Prakash Talgulkar | 16m |
Image mode (On-Cluster Layering) treats RHCOS as a buildable image inside the cluster, using a MachineOSConfig-driven build controller and build jobs to… | ||
| Democratized Deception: How Threat Actors Are Weaponizing AI to Scale Cybercrime | Deepak Koul, Deep Yogeshbhai Gandhi | 36m |
A finance officer at Arup wires 25 million after joining an AI-generated video call that clones faces and voices; attackers can build convincing voices from… | ||
| Don’t Trust Opaque Clouds, Cryptographically Verify Instead! | Chris Butler, Pradipta Banerjee | 40m |
Confidential computing claims hardware-enforced TEEs plus remote attestation let workloads prove integrity on demand; the demo boots confidential VMs on Azure… | ||
| AI-Powered Issue Triage: From Chaos to Clarity in Seconds | Tanwi Geetika | 19m |
An open‑source triage pipeline that runs inside GitHub Actions takes a Repomix snapshot of the repo, calls Gemini 2.0 Flash, and returns structured reports… | ||
| Corporate Success: The Unwritten Curriculum | Nimisha Mukherjee | 40m |
Practical checklist of hard and soft skills for early-tenure engineers: prioritize business value, trim technical debt, and follow Agile practices; use the… | ||
| Fix More Than Code: How Code Reviews Improve Teams, Skills, and Products | Manish Bainsla | 14m |
Argues that code review is a Human API where human feedback transfers institutional knowledge AI cannot, and shows concrete effects: AI removes syntax noise so… | ||
| Sustainable Coding: Why Your "Hello World" Matters | Angad Tewari | 17m |
A tiny Python loop that pings a server every 500 ms and a for loop that reallocates a list each iteration are shown to multiply waste at scale, costing CPU… | ||
| Building Real-Time Control Software: Lessons from Hyperloop | Atharva Arbat | 46m |
Team built a base station that achieves 11ms latency end-to-end between pod and GUI by moving from Python to Rust + Qt + React, using UART to the MVCU and… | ||
| Pod-to-Pod Ping: The Hidden Wiring of Kubernetes | Vibhor Chinda | 17m |
Traces a single ping from container to container, showing concrete steps: containers in a pod share single IP per pod and differentiate by ports; traffic… | ||
| Getting Started with Open-Source Contributions: A Beginner's Guide | Jayapriya Pai, S Ashwin | 18m |
Based on GSoC 2024 and 2025 mentoring experience, the session shows how contributors start by forking, cloning, creating a branch, writing an atomic PR with a… | ||
| Tech for Good- Open Source Solution for Climate and Healthcare. | Rahul Belokar, Priya R Belokar | 23m |
Open-source projects pair satellite and ground data with ML to cut emissions and improve care: OpenClimateFix uses Python/PyTorch on GCP and OpenShift to… | ||
| Understanding the Linux Block Layer-From Single Queue to Multi-Queue Architecture | Nitin U. Yewale | 42m |
Maps a real cp read/write through VFS, page cache, file system and device mapper into the block layer and shows why the legacy single-request Q stalls under… | ||
| Memory-Efficient AI: How PEFT and PyTorch Enable Accessible LLM Fine-Tuning | Parshant Sharma | 12m |
Full fine tuning of a 7B model needs roughly 112 GB of GPU RAM and multi‑GB checkpoints per task, while PEFT trains only 0.1–2% of parameters and cuts memory… | ||
| What Did My Cluster Just Do? Event Intelligence with LightRAG | Mitali Bhalla | 10m |
LightRAG converts short‑lived Kubernetes events into compact embeddings, stores them in a local vector database, and answers natural‑language queries via a… | ||
| Why Your RAG is Failing , And How Open Source Can Fix It | Anindita Sinha Banerjee, Nitish Singh | 27m |
Pipeline shows that messy PDFs — broken reading order, nested tables, images and lost metadata — cause RAG failures long before the LLM. A preprocessing stack… | ||
| Brain Tumor Detection and Classification Using Deep Learning Models | Swaraj Sontakke | 15m |
Automated system trains VGG-16, VGG-19 and EfficientNet-B1 (~94% accuracy) on 7,023 MRI images with resizing, normalization and augmentation, using Adam… | ||
| Agentic AI and Model Context Protocol's Security Vulnerabilities | Mohit Sewak | 51m |
Claims that MCP-connected agents allow injection that over 50% of agentic tasks are vulnerable (InjecAgent benchmark), demonstrate Indirect Prompt Injection… | ||
| Quantization at the Edge: Making a 4GB Model Run on 1GB RAM | Sneha Singh | 15m |
A practical recipe for running large LLMs on tiny ARM boards: offline quantization plus operator fusion, KV-cache trimming, and runtime memory pooling compress… | ||
| ModelPack: Packaging ML models as OCI artifacts made easy | Avinash Singh, Andrew Block | 41m |
ModelPack packages real ML models as OCI artifacts, using a modl CLI to generate a model manifest (similar to a lightweight Dockerfile), build and push layered… | ||
| AI for Legal : The story of hype, adoption and what comes next | Ashwin Vinay Phadke | 46m |
Generative AI accelerates legal work from TAR-style classification to drafting and research, with retrieval setups like retrieval-augmented generation hooked… | ||
| Writing, running, and testing awesome Ansible content with natural language and AI | Shatakshi Mishra | 26m |
A Model Context Protocol server embeds into the Ansible VS Code extension to keep playbooks, roles and inventories in one place and remove tool hopping. Demo… | ||
| Centralized CI Monitoring with GitHub Actions | Anushka Shukla | 15m |
Built a centralized CI observability system that ties 15+ repos into one channel using GitHub Actions reusable workflows, a single central repo for the… | ||
| Scaling Generative AI Inference with llm-d | Dasharath Masirkar | 32m |
llm-d is a Kubernetes-native serving stack that splits inference into prefill, decode and shared KV cache microservices so prefill-heavy work and memory-bound… | ||
| Empowering the Future of ML: Scaling the Kubeflow Community in 2026 | Amita Sharma, Gaurav Kamathe | 17m |
Red Hat lays out a 2026 roadmap to grow the Kubeflow community, recruit contributors at the conference booth, and push upstream collaboration with concrete… | ||
| Running Multi-Arch Builds & Tests in OpenShift CI/CD | Nick Moraitis | 38m |
Operates a CI/CD platform across 11 OpenShift clusters with ~100,000 job definitions and ~500,000 daily executions; implements PROW + CI Operator plus a custom… | ||
| Architecting Efficient Agents with Semantic Routing and Mixture-of-Tools | Vincent Caldeira | 46m |
Claims a semantic router (VLM envoy proxy + small classifier + keyword/embedding signals) routes queries to the right model and cuts cost and time—benchmarks… | ||
| Unpacking the Llama Stack: Architecting Next-Gen AI Applications | Aditya Patil | 17m |
Llama Stack is a production backend not a model: it exposes inference API and endpoints like /chat/completions, /models, /agents and rag/documents, plus… | ||
| A Quantum Computing Talk Without Qubits | Ram Iyengar | 38m |
RHEL 10 and Fedora already package NIST‑approved post‑quantum algorithms — ML-KEM and ML-DSA — into OpenSSL/GnuTLS and a single crypto‑policy switch toggles a… | ||
| MLOps is the New DevOps: How Sysadmins are the Next AI Operators | Hemant Wadhwani | 17m |
Operations skills map to MLOps by handling the artifact lifecycle not the math: take a Jupyter output like model.pkl, wrap it with FastAPI, containerize, push… | ||
| Keynote: Building India’s Sovereign AI Stack w/ Local Talent & Open Global Collab | Vincent Caldeira | 24m |
India builds a DPI-style AI stack that actually ships resources and models: 38,000 GPUs at ~70% discount for startups and researchers, a curated data portal AI… | ||
| Collaboration Synthesis:Engineering India’s Open AI Future from Classroom to Cloud | Ameeta Roy, Balaji Patil, Chris Bredesen, Brian Preffitt, Gaurav Hirani | 48m |
Panelists argue India must move from AI consumers to builders, focusing on open source, smaller local models and production readiness. They cite 38,000+ GPUs… | ||
| Supercharge Your GitOps with ArgoCD Agent | Anand Kumar Singh, Akhil Nittala | 42m |
ArgoCD Agent replaces a single hub reconcilier with per-cluster agents so credentials stay local, reconcilers run inside each workload cluster, and the hub… | ||
| Scaling ML Pipelines with Feast, Ray and Kubeflow | Nikhil Kathole, Abhijeet Dhumal | 46m |
Built a file-based pipeline into a distributed stack using Feast for feature management, Ray as the batch engine, and Kubeflow Training Operator to run PyTorch… | ||
| The Compute Revolution You’re Ignoring: JavaScript in Science | Gunj Joshi | 43m |
Claim: modern JavaScript plus stdlib runs real scientific workloads inside the browser, cutting latency and installs. Evidence: three implementations of an… | ||
| Self-Healing Pipelines? Resilient Deployments with Hardened Containers | Rajani Ekunde | 45m |
Claim: build self-healing CI/CD pipelines that block risky artifacts and automate rollbacks by shifting security left and hardening images. The demo uses… | ||
| The Open Source Community Playbook: What Works, What Doesn’t, and Why | Pritesh Kiri | 16m |
Practical playbook: run LitmusChaos (CNCF incubated) and ReactPlay as examples, run monthly community calls and enforce a monthly release cadence to show… | ||
| FusionStack: The Cross-Platform Blueprint | Samson Dmello, Mandar Dixit | 38m |
Blueprint unifies OpenShift Virtualization (KubeVirt) on‑prem with ROSA on AWS, uses Ansible Automation Platform and ACM to declaratively orchestrate VM… | ||
| From IaC to InfraOps: Automating Day-2 Operations with Terraform Actions & Ansible | Rahul Gaikwad | 37m |
Demo shows a Git push to Terraform Cloud triggers a workspace run that creates two VMs on Google Cloud (5 → 7), calls HashiCorp Vault for credentials, and… | ||
| Sealing Kubernetes with Confidential Clusters | Nitesh Narayan Lal | 17m |
Describes a working Confidential Cluster Operator that automates node onboarding by issuing a UUID per node, calculating and storing baseline measurements, and… | ||
| Art of Innovation: The Artist’s Guide to Product Management | Sandeep Joshi | 16m |
Claims that product work shifts from building UIs to orchestrating AI agents, backed by hands-on use of Google AI platforms alongside Claude and recent ChatGPT… | ||
| LogAn: Large-scale Log Analysis with Small Language Models | Rahul Shetty, Siddardh R A | 43m |
LogAn analyzes GB-scale log dumps by first applying log templatization (Drain algorithm) to collapse 38K lines into a dozen templates, then runs CPU-friendly… | ||
| Resolving Performance and Security Issues with eBPF | Kashyap Vasant Ekbote, Yogesh Bhalchandra Babar | 42m |
eBPF is shown as a kernel‑embedded VM that runs verified bytecode so monitoring hooks can catch ultrafast process creation that top misses; the demo uses execs… | ||
| Apples, Oranges, and ML Models: Model Validation vs Benchmarking | Gaurav Kamathe | 19m |
Red Hat is building a scalable system to evaluate LLMs, distinguishing benchmarking (relative leaderboards like MMLU) from validation (fit for a specific use… | ||
| Think Like an Attacker: Why Your npm install Is a Hacker's Dream | Sudhanshu Dasgupta | 29m |
Shows how a dormant npm package with a suddenly added post-install script allows attackers to exfiltrate npm/GitHub tokens and AWS keys, then auto-bump and… | ||
| The Power of Edge AI: Revolutionizing Future of Real-Time, Autonomous Processing | Deepak Das | 48m |
Edge AI runs inference on-device using small models (1–7B parameters or tiny models in millions), with quantified wins: 100 ms to 10 ms latency in an… | ||
| eSim: Building an Open-Source EDA Ecosystem for Edu, R&D, and Community Innovation | Sumanto Kar | 28m |
eSim packages ngspice, KiCad, GHDL, Verilator and Python/PyQt5 tooling into a usable EDA chain, supports mix-signal simulation, PCB auto-routing and device… | ||
| eBPF-Driven Security for Kubernetes: The Tetragon Story | Aftab S | 31m |
Tetragon uses eBPF to hook into the kernel and capture every process, file and network event with microsecond granularity, writing events to eBPF maps for… | ||
| Unifying AI experience for customers: Building multi-agent systems with Google ADK | Anuj Singla, Omkar Pavale, Deepak Koul | 45m |
Built a supervisor “super-agent” that routes to specialist sub-agents using Google ADK and the A2A protocol, with a live POC demo showing multilingual… | ||
| Simplifying Storage in Kubernetes with Rook-Ceph | Subham Kumar Rai, Nikhil Ladha, Rakshith R | 29m |
Rook bundles Ceph into Kubernetes as a production-ready operator (Rook v1.19 released Jan 2026) that runs on K8s 1.30–1.35 and talks to Ceph 19+, supporting… | ||
| Shift-Left for LLMs: Securing the AI Model Supply Chain | Nagesh Rathod | 36m |
A Kubeflow pipeline fetches a 0.5B model from Hugging Face, stores artifacts in MinIO, applies Sigstore and Cosign signing, runs NVIDIA Garak for… | ||
| Reproducible, Immutable, Bootable: Exploring bootc with Podman Desktop | Praveen Kumar | 41m |
Claim: bootc turns an OCI image into a reproducible, immutable, bootable OS and Podman Desktop adds a cross‑platform UI + extension that builds, runs and… | ||
| Why Your RAG System Hallucinates: Fixing the Content Segmentation Problem | Sharan Harsoor | 43m |
Argument: naive character-splitting breaks semantic units and causes RAG hallucinations, while semantic-aware segmentation yields concrete gains. The… | ||
| FIREWATCH: Automated JIRA Bug Creation Based on CI Job Configuration | Saumya Vishwakarma | 22m |
Firewatch automates JIRA bug creation from CI jobs by reading a JSON config of per-step failure_rules and success_rules, running as a post-step ref called… | ||
| 35 Fedora releases in 35 minutes: a look back with lessons for any project | Matthew Miller | 47m |
Covers 35 Fedora releases with concrete lessons: how community-driven choices like x86_64 adoption and replacing the init with Upstart then systemd changed the… | ||
| Sovereign AI: India’s Next Big Leap | Pranjal Bathia, Renu Jhamtani | 42m |
India argues for building local models, open-source collaboration, data sovereignty, and hardware capability by citing Aadhaar, UPI, ONDC and the Bhashini and… | ||
| Transforming SBOMs from Compliance Burden to Security Asset | Martin Sikora, Ales Raszka | 34m |
Konflux integrates an automated SBOM lifecycle that uses Mobster to run Syft-like scanners during builds, emit SPDX/CycloneDX SBOMs per OCI image and index… | ||
| Breaking the Build Before It Breaks You: The Magic of E2E Tests | Rizwana Naaz, Alka Kumari | 34m |
Demo shows a simple Go todo app broken by an intentional change that unit and integration tests both miss but an end‑to‑end run fails, proving E2E catches… | ||
| Solving the ML Pain You Forgot to Mention | Aniket Paluskar | 34m |
Feast presents an open source feature store that claims to fix training serving skew and point in time correctness by unifying feature definitions, versioning… | ||
| Securing Agentic Platforms with Zero-Trust Workload Identities | Trilok Geer, Andrew Block, Anjali Telang | 36m |
Shows a working pattern: workload attestation with SPIFFE/SPIRE agents on every node, Keycloak as the OIDC/OAuth2 broker (new Keycloak 26.2 SPIFFE identity… | ||
| From Side Project to Industry Standard: Story of AsyncAPI | Shriya Chauhan | 32m |
AsyncAPI claims a community-built spec that reached a $33M valuation and CNCF backing, solves “contract blindness” by describing producers, channels, messages… | ||
| Will AI Replace Writers? No, But It Will Replace Writers Who Don’t Evolve & Adapt | Yash Guddeti, Kalyani Desai | 26m |
Claim: AI handles first drafts, restructuring and summarizing but cannot do the 90% of technical writing that is negotiation, judgment and information… | ||
| AIOps for Distributed Environments - Deep Dive | Andreas Spanner, PRATHEEBA RAVINDRAN | 42m |
Linux Foundation AIOps project builds a test harness and demos that run on OpenShift, ingest OpenTelemetry (metrics, traces, logs) into Kafka/OpenSearch, and… | ||
| Enabling Data Sovereignty in AI Workflows through Data Mesh Architecture | Jeyaramachandran Paulraj, PRATHEEBA RAVINDRAN | 45m |
Claims that a Kubernetes-native Data Mesh delivers data sovereignty for AI by keeping data local while enabling federated queries and model training; uses… | ||
| The GPU Utilization Problem: What’s Going Wrong and How to Solve It | Amita Sharma, Shamsher Ansari | 43m |
Claim: typical data-center GPUs sit at 20–30% utilization now while the goal is 80–90%. Root causes: coarse Kubernetes GPU allocation (nvidia.com/GPU1)… | ||
| Chandrayaan Mindset for Your AI Career: Build Systems That Survive India | Monika Chahal | 18m |
India needs systems that survive multilingual noisy inputs, low connectivity and cheap hardware rather than ever‑larger models; the practical claim is to shift… | ||
| AI Is Not Magic: The Physics, Math, and Hardware Behind It | Adarsh Dubey | 18m |
Explains that LLMs are linear algebra plus optimization running on specialized silicon: words live as embeddings in 768 dimensions, similarity search uses… | ||
| nothing matches | ||