From MIT to IBM, expediting AI and quantum deployment

Three researchers now at IBM describe how their time with the MIT-IBM Computing Research Lab (formerly the MIT-IBM Watson AI Lab) helped connect their academic work to industry applications, according to an MIT News profile. Srinivasan Arunachalam, Zhang-Wei Hong PhD ’25, and Irene Ko PhD ’24 work in quantum machine learning, reinforcement learning and AI agents, and trustworthy and fair AI, respectively.

Hong, an IBM research staff member with the lab, began his MIT PhD in 2020 in the Department of Electrical Engineering and Computer Science. He worked with EECS Associate Professor Pulkit Agrawal, also a lab principal investigator, on improving value function learning for reinforcement learning in video games. At IBM he investigates test-time training for agents and foundation models and develops infrastructure for an IBM agentic framework for enterprise tasks including chart reading and tool calling for database queries. He said that if successful, such a framework would let a model self-evolve its weights online at deployment time.

Ko, who joined IBM Research after graduating in 2024, said her MIT-IBM-funded PhD collaboration with EECS professor Luca Daniel and IBM Principal Research Scientist Pin-Yu Chen shaped her work on trustworthy AI. Her current project, vLLM Hook, accesses internal model signals such as hidden states or activations for decoding LLMs; a vector acts on transformer modules to analyze safety scores, including likelihood of prompt-injection and hallucination. She developed a lightweight vLLM inference engine plugin framework to program model internals, which could provide significant cost savings over other methods.

Arunachalam, an MIT postdoc in 2018 in Professor Aram Harrow’s physics group, collaborated with the lab and IBM researcher Kristan Temme after conversations with Isaac Chuang, an MIT-IBM principal investigator. His work on Hamiltonian learning gave rigorous guarantees for learning quantum system dynamics, and a paper on quantum kernels offered theoretical evidence that quantum feature spaces can offer advantages over classical kernels under widely believed hardness assumptions.

The lab, the profile says, served as a conduit for research relationship building and the flow of expertise to industry applications.