Senior AI Research Scientist

The Reality Check

Most companies say they're "AI-first." We actually mean it. We're building AI platforms that don't just demo well—they ship.

What You'll Actually Do (Not Corporate Fluff)

• Train LLMs from scratch (yes, actually from scratch—not just fine-tuning someone else's model)
• Architect tokenization strategies that don't make other engineers cry
• Push the boundaries of what's possible with transformer architectures
• Debate whether RLHF is overhyped (spoiler: we have opinions, but we want yours too)
• Build models that work in production, not just in Jupyter notebooks
• Occasionally explain to non-technical folks why "just add more parameters" isn't always the answer

You're Our Person If:

✅ You've defended a Ph.D. thesis (and lived to tell the tale)

✅ You've trained LLMs from scratch and know the pain of watching loss curves plateau at 3 AM

✅ Tokenizers aren't black boxes to you—you've built them, broken them, fixed them

✅ You can discuss attention mechanisms, positional encodings, and scaling laws without needing slides

✅ You've shipped models to production (bonus points if they're still running)

✅ You think in tensors but communicate in human

Bonus Points:

🎯 Published at NeurIPS/ICML/ACL/EMNLP

🎯 Contributed to open-source LLM projects

🎯 Have strong opinions about BPE vs. WordPiece (and can defend them)

🎯 Experience with distributed training frameworks

🎯 You've read the "Attention Is All You Need" paper more than once (by choice)

What We Offer:

💰 Competitive salary (because ramen gets old after grad school)

🚀 Actual compute resources (GPUs that aren't shared with 47 other teams)

🧪 Freedom to experiment (and permission to fail fast)

🎯 Impact (your models will be used by Fortune 10 clients, not collecting dust)

🌱 Growth (we're building something big, and you'll shape it)

What We Don't Offer:

❌ Death by meetings

❌ "Agile" theater

❌ Corporate bureaucracy masquerading as process

❌ The phrase "let's take this offline"

The Vibe:

We're technical folks building for technical folks. We care about elegant solutions, clean code, and models that actually work. We don't care about buzzword bingo or AI theater.

Ready to Build the Future?

Drop us a message with:

• Your most interesting LLM war story
• A link to your research
• Why training from scratch excites you (or terrifies you—both are valid)

P.S.