Research Engineer - LLM Post-Training

Research Engineer – LLM Post-Training

We (https://www.kaon.io/) build large-scale LLM systems for roleplay, interactive storytelling, and emotionally intelligent content generation for millions of users. The goal of our post-training team is to turn foundation models into high-quality character agents with strong personality consistency, long-term memory, and immersive narrative ability, while aligning it to online user preferences.

Responsibilities

• Design and implement post-training pipelines for roleplay and generative LLMs, including SFT and reinforcement-based fine-tuning.

• Improve emotional expression, narrative coherence, and character consistency across long multi-turn conversations.

• Build large-scale preference datasets from user interaction logs.

• Design and implement reward systems, including Generative Reward Models (GRM) and Process Reward Models (PRM).

• Apply and extend DPO, GRPO, and related RL methods to optimize contextual consistency, logical correctness, and out-of-character prevention.

• Build evaluation frameworks tied to real product metrics (retention, engagement, satisfaction).

• Run rapid offline experiments and online A/B tests, iterating based on real user signals.

• Work closely with research and product teams to close the loop between model behavior and user experience.

Requirements

• Strong programming skills in Python and/or C/C++ with solid foundations in data structures and algorithms.

• Experience with LLM training, post-training, and reinforcement learning.

• Experience in roleplay systems, virtual character generation, or emotionally intelligent AI preferred.

• Self-driven, hands-on, and comfortable owning problems from research to production deployment.

Compensation

• $200,000 – $500,000 total compensation (base + equity), depending on experience and impact.