Research
I am researching how to make AI agents truly autonomous — not just reactive, but capable of learning from their own experience and improving over time. The focus is on memory, self-correction, and long-horizon reasoning.
Agent Memory & Continuous Learning
Making agents that learn, retain, and self-improve
Researching memory architectures that allow agents to retain knowledge across sessions, learn from past decisions, and continuously improve their own reasoning and execution quality.
01.Research Interests
Agent Memory Systems
Building persistent memory layers that let agents retain context, recall past experiences, and apply learned knowledge to new situations without starting from scratch.
Continuous Self-Improvement
Designing feedback loops where agents evaluate their own outputs, identify failure patterns, and update their behavior over time without human intervention.
Agent Efficiency
Improving latency, tool usage, and decision quality so agents complete long-horizon tasks with higher reliability and fewer failures.
Autonomous Agent Researcher
Designing agents that can generate hypotheses, run experiments, evaluate results, and produce research outputs with minimal human intervention.
02.Applied Research Projects
Autonomous AI Agent Researcher
End-to-end autonomous research assistant
An end-to-end AI research assistant designed to take a problem statement, propose hypotheses, run structured experiments, and produce research outputs. The goal is to give agents the ability to conduct meaningful research independently, retaining what they learn across runs.
Next Direction
The next steps in my research push toward agents that genuinely grow over time:
- →Memory architectures that persist across sessions and surface the right knowledge at the right moment
- →Self-evaluation loops where agents score their own outputs and adjust strategy accordingly
- →Long-horizon agents that can plan, execute, and learn from multi-day tasks autonomously