Presence
A proactive agent that learns reusable workflows and user preferences, deciding when to act, suggest, ask, or stay silent.
Selected work
Ideas into
working systems.
A few projects across agents, vision-language models, and practical evaluation.
Projects
Vision Augmented Language Models
Joint vision-language reasoning with CLIP ViT and StableLM, packaged for reproducible experimentation.
System-1 Decision Models
Benchmarking agent tool selection across 4,000+ decisions and 800+ workflows, measuring accuracy, calibration, latency, and cost.