Karthik Abinav Sankararaman — AI Researcher

I do research on frontier model training — mid/post-training, RLHF, reward modeling, tool use, and agentic behavior. My focus is on making large language models more capable, reliable, and aligned with how people actually use them.At Meta Superintelligence Labs, I lead research on Llama/MetaAI, setting technical direction across data & RL, factuality, model personality & EQ, tool use, and agentic systems. I've developed novel RL algorithms, reward modeling pipelines, and data flywheel systems for continuous model improvement — and worked across teams to translate this research into every Llama release since 2023. Before moving to frontier models, I developed RL and bandit algorithms deployed across Meta's major product surfaces — ads, recommendations, content integrity leading to significant cumulative business impact. This grounded my research in what it means to build systems that work reliably at scale. Along with product impact, I have published several papers covering the algorithmic aspects of these works.I hold a PhD from the University of Maryland in sequential decision making and bandit theory. To know more about me and my research, visit my personal webpage: karthikabinavs.xyz

Stackforce AI infers this person is a leading AI researcher specializing in reinforcement learning and large language models.

Location: San Francisco, California, United States

Experience: 9 yrs 6 mos