🛠 Tech Stack

💼 About This Role

You'll join Distyl's post-training team to adapt foundation models for real-world enterprise performance and alignment. You'll investigate new methods for aligning models with human and system-level objectives, bridging raw capability with trustworthy, contextually aligned system behavior. Your work will impact critical operations across telecom, healthcare, and manufacturing.

🎯 What You'll Do

• Adapt foundation models using supervised fine-tuning and preference optimization
• Develop and evaluate techniques like DPO, RLHF, and continual adaptation
• Align models with enterprise systems for trustworthy behavior
• Explore trade-offs between generalization and specialization

📋 Requirements

• Deep understanding of post-training techniques (SFT, RLHF/DPO, LoRA/PEFT)
• Experience adapting frontier models (LLMs/SLMs) to specialized domains
• Experience building with models, not just building models (compound AI systems, agentic collaboration)
• Strong programming and data analysis skills for prototyping and experimentation

✨ Nice to Have

• Published research in top journals or venues
• Track record of sharing work on platforms like Twitter or Arxiv
• Daily use of AI tools like ChatGPT, Cursor, Perplexity

🎁 Benefits & Perks

• 🏥 100% covered medical, dental, vision for employees and dependents
• 💰 Equity alongside base compensation
• 🏋️ 401(k) with commuter benefits and in-office lunch
• 🤖 Access to state-of-the-art models and generous AI tool usage
• 📈 Ownership of high-impact projects across top enterprises