Independent research collective
We build intelligent systems that remain useful when models hallucinate, networks disappear, and compute is constrained.
01 / OUR POSITION
ResForge is a two-person research team exploring how AI can move from impressive demos to trustworthy, efficient systems.
We combine machine learning, software engineering, and safety-aware system design—then test the whole pipeline, including how it fails.
02 / ACTIVE RESEARCH
CNN-Based Corn Leaf Disease Detection and LLM-Powered Treatment Advisory
Most crop-disease tools stop after classification. MaizeCure connects visual diagnosis to safety-aware treatment guidance—without placing blind trust in a single model.
Four transfer-learned CNN backbones operate independently or together through soft voting and a cost-aware hybrid cascade.
Low-confidence and mixed-infection predictions are filtered before multiple LLM providers validate a recommendation.
Malformed outputs are repaired, terminology is screened, and deterministic expert rules take over if consensus or connectivity fails.
03 / THE NEXT FRONTIER
Our next research chapter focuses on cognitive AI and efficient intelligence that lives on the device—not only in the cloud.
Systems that perceive, reason, adapt, and interact with richer context.
Private, low-latency language intelligence under real hardware constraints.
Model and runtime optimization for the next generation of neural processors.
04 / THE TEAM
Two researchers, one shared standard: build with rigor, question the easy answer, and make the system work beyond the lab.
05 / COLLABORATE