Rekous

Building AI Driven HPC Teams

As an HPC Manager, I’m always thinking about leadership … and when I say leadership, I don’t just mean hitting productivity targets. I mean building a team that is effective, aligned, motivated, and seen as credible by the organization.

In high-performance computing, results matter. But so does presence. The way your team operates, communicates, and presents its work directly reflects your leadership.

Over the last four years, I’ve built a team I’m incredibly proud of. And I don’t think that success has been accidental.

Creating Opportunities, Not Just Assignments

Throughout my tenure, I’ve placed strong emphasis on giving my team ownership of unique, high-impact projects; whether that’s internal tooling or new product features.

I spend a significant amount of time outside of normal working hours thinking about:

This isn’t just about delegation.

It’s about:

I see it as a fundamental responsibility of a manager: continuously thinking about future projects the team can create that elevate both productivity and individual growth.

This approach has been a cornerstone of our success.

AI is Not Hype, but an opportunity for Leverage

Recently, I’ve been diving deeper into AI.

And let me be clear: this isn’t another post about AI “taking over the world.”

But if you’re only using ChatGPT as a prompt-response tool, you’re likely missing the real opportunity.

For example, I’ve built local MCP servers that:

Instead of manually gathering data, I run a prompt. My system fetches, aggregates, formats, and delivers.

That’s leverage.

In HPC environments, where efficiency, automation, and scale are everything. This kind of tooling becomes transformative.

Curiosity Is Still the Root

No matter how advanced the technology becomes, I always return to one core principle:

Curiosity.

If I want to be articulate about something, I need to understand the fundamentals.

Lately, I’ve been studying Github projects like OpenClaw and broader AI agent architectures. There’s a lot of discussion about where AI is heading, but what interests me most isn’t the headlines. It’s the mechanics.

At a high level, agent-based systems (in particular OpenClaw) typically include:

1. Channel Adapters

The messaging interfaces (Slack, Discord, Telegram, iMessage, etc.)

2. Gateway

The routing layer that handles sessions, queues, and concurrency.

3. Agent Runner

The orchestration engine that:

The Agent Runner is the most interesting component to me. It’s the brain of the operation.

Up until now, I’ve mostly interacted with LLMs through Copilot CLI or VS Code integrations. But I’ve never built the orchestration layer myself.

So that’s likely my next project:

Build my own Agent Runner from scratch.

Not because I need to.
But because understanding it will reveal gaps, design trade-offs, and maybe even opportunities for a unique twist tailored to HPC workflows.

Leadership Through Understanding

As a manager, I believe credibility matters and more importantly:

Credibility is earned.

If I’m going to encourage my team to explore AI, automation, or next-generation tooling, I need to understand it at a fundamental level myself.

That’s part of being an example.

The future of HPC teams won’t just be about clusters, networking, RDMA, or scheduling efficiency. It will be about:

The leaders who thrive will be the ones who:

For me, leadership isn’t about control.
It’s about positioning the team — and each individual — to win together.

And in the age of AI, that positioning requires both vision and technical curiosity.

I’m excited to see where it leads.

#Technology