Fern GTM brain

Fern builds and hosts high-fidelity robot simulation environments for robotics teams training and evaluating policies.

category

Robotics simulation and reinforcement learning infrastructure

market profile

  • text: Physical robot evaluation consumes operator time and is difficult to reproduce across policy checkpoints.
  • text: Robotics teams can build simulation workflows using open frameworks such as NVIDIA Isaac Lab, making internal build versus external service a relevant choice.

personas

  • title: Head of robot learning
    pains:
    • Policy evaluation on hardware is slow and conditions vary between runs.
    • Training at scale requires many rollouts without monopolizing physical robots.
  • title: Robotics simulation or infrastructure lead
    pains:
    • Matching geometry, contacts, cameras and controllers to physical equipment takes tuning.
    • Maintaining simulation clusters and changing production layouts draws engineers away from policy work.
  • title: Robotics startup CTO
    pains:
    • Limited hardware and engineering capacity slow experimentation.
    • A simulator that fails to match real behavior can mislead investment in a policy.

icp

  • text: Robotics companies developing learned manipulation policies and evaluating performance against physical hardware.
  • text: Teams with real robot data and repeated evaluation or reinforcement learning workloads, especially those lacking dedicated simulation infrastructure.

message pillars

  • text: Scale reinforcement learning and evaluation beyond physical hardware capacity.
  • text: Reduce the work of operating simulation infrastructure.
  • text: Ground environments in the customer's physical robot data.
  • text: Free robotics teams to improve policies.

objections

  • objection: Will simulated performance predict our real robot performance?
    response: Test held-out policies and scenarios against physical runs; Fern claims very little sim-to-real gap, which should be validated per task.
  • objection: Why not use an open simulator internally?
    response: Compare total cost of tuning, hosting and maintaining task-specific environments with a bounded external pilot.
  • objection: Can the service handle our robot and production changes?
    response: Review supported data and interfaces and measure turnaround on a representative layout change.

competitors

  • name: NVIDIA
    website: https://developer.nvidia.com/isaac/lab?ncid=so-yout-329964
    why: Isaac Lab and Isaac Sim provide an alternative stack for robot learning and simulation.
  • name: Google DeepMind
    why: MuJoCo is an open physics engine that teams can use to build robotics simulation workflows.

keywords

  • robotics simulation environments
  • robot manipulation reinforcement learning
  • robot policy evaluation
  • sim to real robotics
  • robot learning simulation
  • parallel robot rollouts

voice

  • Technical and research-oriented
  • Direct and outcome-focused

signals

  • text: A robotics company is hiring robot learning or simulation engineers and running manipulation policy experiments.
  • text: Policy testing is constrained by access to real robots or by maintaining a custom simulator.

pricing

model: unknown

motion

sales_led

channels

  • technical whitepaper
  • LinkedIn
  • website meetings

confidence

medium

Founders

  • name: Amit Yadav
    title: Founder/CEO
    role: CEO
    bio: Cofounder and CEO at Ishiki Labs (W26). Previously: Research Scientist, first in LlaMA team training multimodal LLMs and then in Reality Labs at Meta training video assistant for smart glasses. PhD from Purdue University with 20+ publications at top conferences like CVPR, NeurIPS, and ICASSP
  • name: Robert Xu
    title: Founder
    role: Founder
    bio: Co-founder & CTO of Ishiki Labs (W26). Previously worked on multi-modal AI and Orion AR glasses at Meta and research infra at Citadel Securities.

Back to YC companies