Tashi Tengyal

I broadly work on/with LLMs for test-time scaling. My core interests are reasoning, memory and decision-making in LLM agents.

At Rezi, I work on multi-agent systems, retrieval and recsys for job applications and career workflows. I primarily use Python with Langchain, DSPy and ElasticSearch at work.

On the side, I like conducting toy experiments (often reproduction or tangential) mainly for test-time scaling methods. Lately, I have been spending some time in RL for post-training. I document these experiments in my blogs.