[03/29/26 11:56:40] INFO scene_file_writer.py:904 File ready at '/Users/abhishekgupta/Documents/Git/PitchDeck/media/images/Ma rketSizing_ManimCE_v0.20.1.png'
INFO Rendered MarketSizing scene.py:278 Played 0 animations
Manim Community v0.20.1
Ensemble Autonomy
Pitch Deck
Abhishek Gupta
Associate Professor, ECE, The Ohio State University
Founder, Rewardwise & Ensemble Control Inc.
Ensemble Robotics
Fault Tolerant Robotics
Abhishek Gupta, PhD
Associate Professor, The Ohio State University
ex-UIUC, USC, UC Berkeley, Stanford
We make Physical AI composable and fault-tolerant
The Problem: The “Loner” Robot Gap
- Current State: General purpose robots are built for one class of tasks
- The Industry 2.0 Challenge: We need different types of General Purpose Robots (humanoid, dogs, snakes) that work as a team
- Daily Complexity: Missions change every day; the system must handle different tasks without manual reprogramming
| From: Rigid & Fragile | To: Adaptive & Resilient |
|---|---|
| Standalone Operations | Coordinated Collaboration |
| Fixed Routines | High-Variety Missions |
| System Stops on Error | Self-Healing Workflows |
| Manual Intervention on Error | Auto-reconfigure for Mission Objectives |
Solution: A Brain That Coordinates & Acts
We are building a Hierarchical Multimodal-Language-Orchestration (MLO) Model. Think of it as the Air Traffic Control for physical AI:
Multimodal: It captures all the information from robots in the field.
Language: It understands your instructions and mission objectives.
Orchestration: It determines an orchestration plan – sequence of tasks and identifying the best robot to carry out that task.
Why This Matters
The “Databricks” for the Physical World
Just as Databricks organizes massive amounts of digital data, we organize physical work.
Time to Value: No need to spend months coding. The system is “ready to work” out of the box.
Plug-and-Play Collaboration (Composable)
Adding a new robot is like adding a new worker to a crew, with skills shared autonomously.
No “manager” needed; the system seamlessly integrates them to tackle bigger tasks together.
Battle-Tested (Robustness and Fault Tolerant)
- Environmental Degradation: Whether it’s a dark warehouse, a dusty construction site, or wifi deadzones, the system doesn’t quit. It stays focused on the mission even when conditions get messy.
Market Size
Competition
Manim Community v0.20.1
Product
A Databricks type coordination platform that
breaks down mission into tasks
divides the tasks to the robots in a sequence
monitors for error
takes preemptive actions to restrict the cascading effects of the error
logs data for troubleshooting and future training
to achieve the mission objectives
Business Model
Partner with multiple robotics companies and provide them with a simulation model to check for the integration effort needed
Multi-robot coordination is where failure happens
- Communication bottlenecks and dropouts
- Deadzones in 10000 sq. ft warehouse
The industry is obsessed with ‘World Models’ and planning for single general-purpose robot. To make it fault resilient, we need an orchestration layer with guaranteed performance.
Technical Moat
My research group has developed many algorithms for multi-agent coordination, which can be productized for robotics environment
Seeking $50M for seed round for 20% stake
| Category | Allocation | Budget | Strategic Objective |
|---|---|---|---|
| Compute & R&D | 40% | $20.0M | GPU Clusters (H100/B200) & Simulation |
| Engineering Talent | 35% | $17.5M | Scaling to 20+ RL & Robotics Leads |
| Hardware & Labs | 15% | $7.5M | Prototype Iteration & Columbus Lab Ops |
| GTM & IP Strategy | 5% | $2.5M | Patent Moat & Early Customer Pilots |
| Operations Reserve | 5% | $2.5M | Scaling Buffer & Contingency |
| Total | 100% | $50.0M | 24-Month Runway to Series A |
Abhishek Gupta | Associate Professor, OSU | Founder, Ensemble Robotics