We’re a science-driven firm, applying the rigor of perceptual and cognitive research to AI enterprises.
Perceptual Insights Inc. (piinc) is founded by Zoya Bylinskii and Aude Oliva.
We bring decades of combined expertise in human perception, cognition, neuroscience, and AI, uniquely blending pioneering AI research with hands-on industry experience in large-scale generative AI evaluation.
Pain: Evaluating multimodal AI is nuanced and challenging.
Find your unique model differentiators and detect model failures through rigorous human evaluation, supported by our perceptually-grounded evaluation strategy and
multimodal experts.
Pain: AI models perform well on benchmarks but fail in real-world use.
Select the best AI models for your product through independent benchmarking and human-aligned evaluation, by measuring how models perform on the tasks people actually care about.
Pain: Human annotation pipelines are expensive and inefficient.
piinc serves as a translation layer between human perception & cognition and the real-world development of AI systems. We target measurable improvements in model quality, multimodal evaluation, and data annotation efficiency—increasingly critical bottlenecks for frontier AI labs and enterprises.
Scientifically validate AI performance to diagnose model capabilities and failure modes, targeting interventions with measurable impact.
Leverage perceptual research to optimize data annotation workflows, multimodal evaluation, and human-in-the-loop AI systems.
Equip enterprises with rigor and confidence to make forward-looking investments in generative AI through independent technical diligence.
We collaborate with a Scientific Advisory Network of leading academics and frontier researchers in perception, cognition, and human-centered AI, covering the multimodal spaces of vision, audio, tactile, language, and reasoning. The Insighters bring their scientific perspective and thought partnership to real-world AI development, supporting evaluation strategy, research-informed decision-making, and human-centered model design through opt-in consultations, invited panels, and targeted advisory engagement.
Let’s discuss how we can bring scientific rigor to your AI roadmap, find your differentiators, and help you make confident AI investments.