Socle Partners

Companies building the substrate of intelligence.

We back technical architects constructing enduring moats in workflow orchestration, post-training infrastructure, and synthetic physical environments.

Plena Health
Plena Health↗
COHORT 01 · SEED
CORE WORKFLOW OS
Founded: 2025
Sector: Healthcare & Life Sciences
Focus: Healthcare AI · Enterprise Architecture

“The AI operating system for specialty practice.”

Plena is building the workflow layer that sits above fragmented healthcare IT — connecting EMR, fax, insurance portals, scheduling, billing, and patient communication into a single automated operating system for specialty clinics. Rather than replacing existing software, Plena orchestrates it.

INVESTMENT THESIS
01

True platform value lives in the Workflow Engine, not in any single AI agent.

02

Each new deployment compounds — new integrations, workflow logic, and operational knowledge reduce cost and raise capability for every subsequent clinic.

03

Land-and-Expand from one workflow (referral, phone routing) to full-practice automation unlocks an ACV path from $60K to $240K+.

Goaly
Goaly↗
COHORT 02 · PRE-SEED
POST-TRAINING INFRA
Founders: Ex-Meta Llama Platform & AWS Infrastructure Architects
Founded: 2024
Sector: Frontier Models & AI Infrastructure
Focus: Post-Training & RL · Inference & Serving

“Post-training infrastructure for the enterprise AI era.”

Goaly is building the infrastructure that lets enterprises train proprietary models on their own data — efficiently, continuously, and without a dedicated ML team. As foundation models commoditize, the competitive moat for every enterprise shifts to the models trained on their private data.

INVESTMENT THESIS
01

Foundation models are becoming commodities. The enterprise moat is in proprietary post-trained models, not access to base capabilities.

02

Post-training is the next infrastructure layer after inference — analogous to what Databricks did for data.

03

Exceptional founder-market fit: the team built the systems this product replaces.

Fleet AI
Fleet AI↗
COHORT 03 · SERIES A
AGENT SIMULATION & DATA
Founded: 2024
Sector: Frontier Models & AI Infrastructure
Focus: Data & Simulation · Post-Training & RL

“Training gyms for agents. The data infrastructure of the agent era.”

Fleet builds the environment layer between foundation models and the real world — high-fidelity simulations of enterprise software where agents can train, fail, iterate, and accumulate experience before touching production.

INVESTMENT THESIS
01

Experience Data, not Environment, is the long-term moat.

02

Compounding generalization: each new environment expands the data flywheel, making every subsequent model better at unseen tasks.

03

Fleet is positioned to be the Scale AI of the post-training era — not a data vendor but a system that continuously produces data.

Clair Health
Clair Health↗
COHORT 04 · SEED
CONTINUOUS MOLECULAR DATA
Founded: 2025
Sector: Healthcare & Life Sciences
Focus: Continuous Biosensing

“The world's first noninvasive continuous hormone monitor.”

Clair Health is building the first noninvasive continuous hormone monitor (CHM) — tracking estrogen, progesterone, LH, FSH, and PdG in real-time through wrist-worn biosensors.

INVESTMENT THESIS
01

Continuous molecular telemetry is the next vital sign frontier beyond heart rate and glucose.

02

Proprietary biosensor hardware generates the rarest biometric data asset: longitudinal, high-frequency endocrine trajectories.

03

On-device calibration and hormone intelligence models turn raw sensor signals into actionable physiological guidance.

Generalist
Generalist↗
COHORT 05 · SERIES A
Embodied AI & Robotics
Founded: 2026
Sector: Frontier Models & AI Infrastructure
Focus: Frontier Models / LLMs · Data & Simulation

“Building general intelligence for the physical world.”

Generalist AI is a frontier AI research and product-driven company focused on building general intelligence for the physical world. Founded in 2024, the company trains foundational embodied AI models (such as their GEN-1 and GEN-1.5 series) that allow multi-purpose robots to learn new physical tasks autonomously or adapt rapidly across diverse hardware and end-effectors.

INVESTMENT THESIS
01

Embodied Foundation Models: Building general-purpose robot intelligence that can transfer across tasks, environments, and hardware, targeting a foundational layer for the emerging Physical AI ecosystem.

02

Scalable Robot Learning: Large-scale physical interaction data enables models to generalize and rapidly adapt to new tasks, with GEN-1.5 demonstrating in-context learning from as little as 12 seconds of human demonstration.

03

World-Class Founding Team: Combines frontier AI research and real-world robotics expertise from Google DeepMind, OpenAI, and Boston Dynamics, positioning the company to tackle the technical challenges of general-purpose robotic intelligence.