Perceptron ML
Perceptron ML builds bespoke AI systems that make law firms AI-native.
Competitive Advantages
Perceptron ML differentiates on verified accuracy and private deployment: its grounding engine traces every fact to a real document and verifies it against the primary source before a model may use it. Systems train on each firm’s own matters and run inside the firm’s environment instead of shared off-the-shelf legal AI.
Independent trade coverage notes this taps the appetite for firm-controlled AI sovereignty, singling out Perceptron among the YC legal cohort for likely interest from larger firms. The shared platform approach also lets the two-person team build broadly without operating as an agency.
Source: artificiallawyer.com
Market Outlook
Perceptron ML operates as a two-person founding team in Y Combinator’s Summer 2026 batch with Active status, headquartered in San Francisco. The founders pair NVIDIA AI Research experience with NASA-supported research leadership from their prior academic and venture work.
Early traction is programmatic: acceptance into YC S26, a founder-authored August 2026 launch, and independent trade press naming Perceptron the cohort legal-tech entry most likely to draw large-firm interest. The company solicits managing-partner introductions as it converts launch attention into firm deployments.
Source: ycombinator.com
Memo
Perceptron ML is a San Francisco AI company founded in 2026 by brothers Michael Marcotte and Peyton Marcotte, backed by Y Combinator in the Summer 2026 batch. The company builds bespoke AI systems for law firms, trained on each firm’s own matters and deployed privately inside the firm’s environment.
Its platform centers on a grounding engine that verifies every fact against a primary source before a model may cite it, paired with connectors into firm systems and an agent runtime covering timekeeping, intake, research, discovery, and drafting.
Source: perceptronml.com
Product Overview
Perceptron ML offers bespoke AI systems built around how each law firm already works, rather than an off-the-shelf product.
Each engagement reuses one platform: the grounding engine, connectors into firm systems, and an agent runtime. Work ships as firm-specific systems such as 2 a.m. intake answering, demands drafted from the record, and deadline and statute monitoring.
Source: perceptronml.com