LLM Researcher

Division: DATUM, Impac Exploration Services
Location: Remote, Oklahoma City (OK), Houston (TX)
Type: Full-Time

About the Role

We're not hiring you to fine-tune prompts. We're hiring you to obsolete them.

DATUM is building a small team of ML researchers who think bigger than today's architectures. We're the ML division of a 40-year-old energy company with startup DNA and the resources to match our ambitions. While others debate scaling laws, we're questioning the fundamentals.

Here's the deal: We have compute. We have some of the world's most complex operational data. We have domain experts who've been modeling physical systems since before transformers were invented. What we need is someone who sees opportunity where others see constraints.

This role is for researchers who read papers and think "that's clever, but here's what they missed." Who understand why current architectures fail on certain problems—and have ideas about what comes next. Who can ship code on Monday and submit to NeurIPS on Friday.

Fair warning: Do exceptional work here and you'll probably get poached. When OpenAI or Anthropic comes calling with their offer, we'll write your recommendation letter. Building things that make Big Tech nervous is how we measure success.

What You'll Actually Do

• Design and train models that don't exist yet—because current approaches aren't good enough
• Turn energy domain challenges into ML breakthroughs (think: models that understand physics, not just language)
• Prototype fast, fail faster, and ship the stuff that works
• Write papers that matter. Open source what you can. Give talks that get noticed.
• Work directly with operators and engineers who will actually use what you build
• Have genuine freedom to pursue "what if we tried..." ideas with real compute behind them

Requirements

• You've trained models from scratch, not just fine-tuned APIs
• Deep understanding of transformers and what comes after them
• Track record of original thinking in ML (papers, code, or products)
• Python/PyTorch/JAX fluency—you build, not just theorize
• Comfort with ambiguity and ability to find signal in noisy, real-world data
• You explain complex ideas simply and challenge assumptions respectfully
• We are not currently sponsoring visas or participating in CPT programs at this time

Bonus Points

• Published ML research or significant open-source contributions
• Experience with scientific computing, physics-informed ML, or industrial systems
• You've built something that surprised even you
• Comfortable in environments where the roadmap is "let's find out"

Why This Matters

Energy is a trillion-dollar industry running on Excel and intuition. The intersection of ML and energy systems is barely explored. While everyone else fights over consumer chatbots, we're working on problems that actually move the needle on energy independence, efficiency, and human progress.

If you want to optimize ad clicks, this isn't for you. If you want to build models that understand how the physical world works—and ship them to people who operate it—let's talk.