Conference Talk Proposals: Phase 3 Authority Building

Conference Talk Proposals for HPC + DevOps + Kubernetes

Overview

Your unique positioning (10+ years HPC + modern DevOps + Kubernetes expertise) opens doors to several high-impact conference venues. These proposals are designed to establish you as a recognized expert in the HPC+Kubernetes intersection while building authority in the technical community.

Proposal 1: “Running Genomic Workloads on Kubernetes: From Local Dev to Multi-Cloud Production”

Target Conferences:

Duration: 45 minutes (presentation) + 15 minutes (Q&A)

Abstract:

Genomic analysis has entered the cloud era, but the tooling hasn’t caught up. Researchers need pipelines that run identically on laptops, local clusters, and cloud infrastructure—without rebuilding for each platform.

In this talk, I’ll share how we built a production genomic analysis platform on Kubernetes that handles 15,000+ samples/month while reducing infrastructure costs by 60% and enabling researchers to deploy their own pipelines in minutes.

Learn:

Who should attend:

Talk Outline (60 minutes total):

0-5 min: Problem statement: genomic workflows at scale 5-15 min: Why Kubernetes + why Slurm isn’t enough 15-30 min: Architecture deep-dive (Argo + Kueue + Karpenter) 30-40 min: Live demo or walkthrough (submit workflow, watch it scale) 40-50 min: Real-world results and lessons learned 50-60 min: Q&A

Key Takeaway: Kubernetes isn’t just for microservices—it’s the platform for reproducible, portable scientific computing at scale.

Proposal 2: “Taming the Cost Beast: From $50k to $5k—Cost Optimization Patterns for HPC in the Cloud”

Target Conferences:

Duration: 45 minutes + 15 minutes Q&A

Abstract:

Cloud is expensive. HPC workloads are doubly expensive. Most organizations over-provision infrastructure, pay for unused capacity during off-peak hours, and have no visibility into cost attribution.

I’ll show you how we reduced monthly infrastructure costs from $50k to $5k (an 80% reduction) while maintaining 99.5% availability and improving utilization from 18% to 72%.

Learn a multi-layer optimization strategy:

I’ll share the exact tools, YAML configurations, and governance patterns that made this possible—plus the hard-won lessons about tradeoffs between cost and latency.

Who should attend:

Talk Outline (60 minutes total):

0-5 min: Cost problem statement (metrics from real organization) 5-15 min: Cost structure breakdown (what’s actually expensive) 15-30 min: Multi-layer optimization strategy walkthrough 30-45 min: Real configurations and tool recommendations (Karpenter, Kueue, Kubecost) 45-55 min: Results, metrics, and lessons learned 55-60 min: Q&A

Key Takeaway: 80% cloud cost reduction is achievable without sacrificing reliability—you just need the right architecture and governance.

Proposal 3: “Infrastructure-as-Code for HPC: Building Reproducible, Multi-Cloud Deployments”

Target Conferences:

Duration: 45 minutes + 15 minutes Q&A

Abstract:

Infrastructure-as-Code (IaC) is non-negotiable for modern DevOps. But applying IaC principles to HPC environments is non-trivial: multiple clouds, heterogeneous workloads, complex networking requirements.

In this talk, I’ll share a production IaC architecture that enables:

Learn:

Who should attend:

Talk Outline (60 minutes total):

0-5 min: Problem: manual infrastructure deployments (pain points) 5-15 min: IaC philosophy and benefits for HPC 15-30 min: Terraform design patterns (modules, variables, outputs) 30-45 min: Multi-cloud architecture walkthrough + live demo 45-55 min: GitOps workflows and disaster recovery 55-60 min: Q&A

Key Takeaway: Infrastructure-as-Code isn’t just about automation—it’s about reproducibility, auditability, and the ability to scale from one datacenter to many without doubling your operational burden.

Proposal 4: “Migrating from Slurm to Kubernetes: A Practical Guide for HPC Administrators”

Target Conferences:

Duration: 90 minutes (workshop format better than talk)

Abstract:

Kubernetes is coming to HPC. The question isn’t “if” but “when” and “how.”

This workshop demystifies the Kubernetes migration path for HPC teams. We’ll cover practical aspects: how existing Slurm workflows map to Kubernetes, what you gain and lose, and realistic migration timelines.

Topics:

By the end, attendees will have a clear migration roadmap for their own organizations.

Who should attend:

Proposal 5: “Observability for HPC: Monitoring and Debugging Compute Workloads at Scale”

Target Conferences:

Duration: 45 minutes + 15 minutes Q&A

Abstract:

You deployed a Kubernetes cluster for HPC workloads. Great! Now why is your job taking 2x longer than expected? Which node has the bottleneck? Where did your 100 jobs disappear to?

Observability for HPC is different from observability for microservices. You care about:

Learn:

I’ll share observability patterns from monitoring 15,000+ monthly job runs across multiple clusters.

Talk Outline (60 minutes):

0-5 min: Problem statement (why HPC observability is hard) 5-15 min: Metrics that matter (throughput, latency, utilization, cost) 15-30 min: Prometheus setup for HPC workloads 30-45 min: Custom metrics and dashboards 45-55 min: Debugging real HPC performance issues 55-60 min: Q&A

Submission Strategy & Timeline

Phase 1: Immediate (Next 2-4 weeks)

Phase 2: Medium-term (2-3 months)

Phase 3: Long-term (6+ months)

How to Strengthen Proposals

Each proposal should include:

  1. Speaker Bio (50 words): “Senior HPC Architect & DevOps Engineer with 10+ years experience scaling high-performance computing infrastructure. Currently leading genomic analysis platform on Kubernetes processing 15,000+ samples/month across multiple clouds. Open-source contributor (Kubernetes, Kueue). AWS certified solutions architect.”

  2. Talk Type: Technical deep-dive (not introductory)

  3. Target Audience: Be specific (bioinformaticians, HPC admins, DevOps engineers)

  4. Learning Outcomes: What will attendees know/be able to do after your talk?

    • Understand architecture decisions and tradeoffs
    • Know specific tools and configurations to use
    • Avoid common pitfalls based on real-world experience

  5. Difficulty Level: Intermediate to Advanced (not beginner-friendly)

Success Metrics

For 2026 (Year 1):

For 2027 (Year 2):

Next Steps

  1. Choose 2-3 proposals that align with your interests
  2. Research deadlines for target conferences
  3. Customize abstracts for each venue (KubeCon abstract ≠ DevOps Days abstract)
  4. Submit early (deadlines are hard stops)
  5. Track submissions in a spreadsheet (conference, deadline, submitted date, status)
  6. Prepare backup plans (if one doesn’t get accepted, you have others in pipeline)

Pro tip: Don’t wait for “perfect readiness.” Submit to multiple venues. The acceptance rate is typically 20-30%, so you need volume.

Good luck! The HPC + Kubernetes community needs more voices like yours.