Seed Round Open · Rolling Closings Through Q4 2026

Local AI that acts inside real software.

The local runtime layer for agentic software.

Run AI agents inside games, simulations, enterprise tools, robotics systems and mission-critical environments without cloud latency, recurring API costs or loss of control.

  • PRODUCTION-PROVEN
  • PATENT-PENDING
  • $3M SEED

Cloud AI can answer. Pontex lets agents act locally.

  • $3M
    Round Target
  • SAFE
    Post-Money
  • $50M
    Valuation Cap
  • Feb 2026
    Live Production
  • 3
    Core Patent Families
  • 6
    LLM Families
  • 4
    Backends
  • 2
    Environments

02 Problem + Why Now

Local models can think. Real software still needs a way to let them act.

Developers are forced to stitch together inference, memory, speech, retrieval, scheduling and tool execution just to connect AI safely to live software.

Smaller Models

Capable AI is moving onto the device.

Stronger Hardware

GPUs, CPUs and NPUs can run increasingly powerful AI locally.

Demand for Control

Lower latency, privacy, offline operation and reduced cloud dependency.

Running a model locally is not enough. Software still needs an execution layer.

The model layer is becoming local.

The software layer is becoming agentic.

Pontex sits at the intersection.

Pontex is built for that shift.

03 Solution

Pontex embeds AI agents where decisions and actions happen.

AIModels · Speech · Memory · Retrieval
Pontex RuntimeReason · Remember · Route · Validate · Execute
SoftwareGames · Simulations · Enterprise · Robotics
Controlled ActionTools · Workflows · Commands · Environment Changes
  • LOCAL
  • REAL-TIME
  • OFFLINE-CAPABLE
  • HARDWARE-AGNOSTIC
  • HOST-CONTROLLED

Cloud AI can answer. Pontex lets agents act locally.

04 Traction

Not a prototype. Not a pitch.

Pontex is live in production.

PRODUCTION DEPLOYMENT

First production deployment inside the founding team’s TCG World Unity HDRP environment.

Full local agent loop

Voice → Reasoning → Memory → Tool Call → Environment Action → Response

LIVE DEMO AVAILABLE

Request Demo
  • Feb 2026
    Live Production
    TCG World deployment
  • ~0.3 sec
    Voice Latency
    measured local loop
  • ~4 sec
    Tool/Action
    full action cycle
  • 6
    LLM Families
    validated locally
  • 4
    Backends
    CUDA · DirectML · Vulkan · CPU
  • 2
    Environments
    Unity + desktop agent

Voice → Reasoning → Memory → Tool Call → Environment Action → Response

Local execution. No cloud inference required. CUDA · DirectML · Vulkan · CPU · Foundry Local compatible

05 Moat + IP

Models will change. The architecture is the asset.

Core Runtime IP

3 Core Patent Families

Native Inference + Memory Bridge

Local inference and zero-copy/native memory handoff

Real-Time Execution

Scheduling, hardware routing and deterministic replay

Safe Agent Control

Speech, RAG, structured output and controlled tool execution

Defensible infrastructure, not a commodity wrapper.

Structural Moat

Production hardening, cross-stack portability and host-native control create a defensible execution layer.

Defensible by Design

  • ProductionPROVEN
  • Model + HWAGNOSTIC
  • Host-NativeCONTROL

Partners integrate once, preserve model and hardware choice, and avoid rebuilding the execution layer internally.

Compounding Advantage

Built to remain the execution layer as models, hardware and host environments evolve.

06 Category Position

Model serving is not agent execution.

ApplicationGames · Simulation · Enterprise · Robotics
PontexHost-Integrated Agent ExecutionState · Memory · Scheduling · Safe Actions
Model ServingFoundry Local · llama.cpp · Local Runtimes
Models + HardwareOpen Models · GPU · CPU · NPU · Edge
  • Model-Agnostic

    Choose the right model for the deployment.

  • Hardware-Agnostic

    Run across available local compute.

  • Host-Native

    AI operates inside the actual software environment.

Pontex is the host-execution layer for local AI.

07 Business Model

Pilot → License → Integrate → Deploy

  1. 1 · PILOT

    Focused integration in the customer’s environment

  2. 2 · LICENSE

    Commercial runtime / SDK license

  3. 3 · INTEGRATE

    Embedded runtime layer across products

  4. 4 · DEPLOY

    Strategic infrastructure deployment

Revenue Model

  • Product Licensing

    SDK · per-product · per-title

  • Enterprise Deployment

    licensing · support · implementation

  • Strategic Integration

    platform · field-of-use · long-term agreements

Pontex can be licensed as a product, integrated as a platform layer, or deployed as strategic infrastructure.

08 Market Opportunity

A massive market is moving toward local, agentic software.

  • $47B
    Global Edge AI
    2026
  • $446B
    Edge AI Projected
    2034
  • Up to 40%
    Enterprise Apps
    with task-specific AI agents by 2026

Pontex targets the execution layer inside that shift.

  • Gaming + Simulation

    Games · Engines · Digital worlds

  • Enterprise Software

    Private agents · Workflows · Applications

  • Defense + Secure Systems

    Simulation · Air-gapped systems · Mission environments

  • Industrial + Robotics

    Automation · Digital twins · Physical AI

  • Hardware + OEM

    PCs · Edge devices · Embedded platforms

  • Sovereign AI

    Private · Local · Controlled deployments

500+ STRATEGIC LICENSING TARGETSInitial strategic universe across software platforms, game studios, OEMs, defense, industrial and sovereign AI.
Discuss a Partnership

One runtime architecture. Multiple high-value licensing markets.

Sources: Fortune Business Insights, Edge AI Market forecast; Gartner, task-specific AI agents in enterprise applications.

09 Team

We built the technology before we raised the capital.

  • David Evans, Co-Founder & CEO

    David Evans

    Co-Founder & CEO

    Leads strategy, partnerships and commercialization from the production environment where Pontex was first proven.

    LinkedIn, David Evans
  • Justin Del Giudice, Co-Founder & COO/CFO

    Justin Del Giudice

    Co-Founder & COO/CFO

    Leads finance, operations, corporate execution and fundraising for Pontex.

    LinkedIn, Justin Del Giudice
  • Richard Santos, Co-Founder & Technical Inventor

    Richard Santos

    Co-Founder & Technical Inventor

    Architected and shipped the Pontex runtime and leads next-generation runtime research. Inventor across the core runtime IP.

Founder-built. Product-first. Already executing.

Capital accelerates Pontex. It does not create it.

10 The Round

$3M Seed · Post-Money SAFE · $50M Valuation Cap

Scale a proven runtime into a category-defining AI infrastructure platform.

  • $3,000,000
    Round Target
  • SAFE
    Instrument
    post-money
  • $50,000,000
    Valuation Cap
  • Q4 2026
    Rolling Closings

Use of Funds

  • Scale the Platform

    • Commercial SDK/runtime
    • Developer tooling
    • Multi-platform integration
    • Enterprise/security deployment
  • Drive Deployment

    • Design-partner integrations
    • Production licenses
    • Repeatable licensing motion
    • Strategic platform relationships
  • Extend the Moat

    • Core patent prosecution
    • Runtime Architecture IP
    • Production Hardening at Scale
    • Next-generation local AI architecture

18–24 Month Outcomes

  • Commercial SDKdeployed externally
  • Sales Motionproduction licensing
  • Multi-Platformruntime deployments
  • IP Portfolioexpanded protection
  • Enterprise ReadinessSecurity + deployment hardening

We proved we can build it. Now we are raising to scale it.

The companies that control local agent execution will control a critical layer of the next AI stack. Pontex is built for that layer.

11 Contact

Indicate your interest.

Tell us whether you are exploring an investment, would like a private demo, or want to discuss a strategic partnership. The Pontex team will follow up directly.

Investment Materials

Investor DeckPDF · 10 pages

Contact Directly

partnerships@pontex.ai

pontex.ai

Accredited investor acknowledgements

By submitting, you agree to receive follow-up communications from Pontex regarding your inquiry. Do not send financial or accreditation documents through this form; accredited investor verification is handled separately through a third-party service. We handle your details as described in our Privacy Notice.

Frequently asked questions

Pontex is the local runtime layer for agentic software. It lets AI agents operate directly inside games, simulations, enterprise tools, robotics systems and mission-critical environments without cloud latency, recurring API costs or loss of control.

Pontex is raising a $3M seed round through a post-money SAFE with a $50M valuation cap. The offering is being conducted under Rule 506(c) and investment is limited to verified accredited investors.

There is no fixed minimum stated for this round. Please indicate your estimated range in the form and our team will follow up directly.

Pontex is currently pre-revenue. Its first production deployment has been live inside TCG World’s Unity HDRP environment since February 2026.

Capital will be used to scale the commercial SDK/runtime, expand developer tooling and multi-platform integration, drive production licensing and strategic deployments, harden enterprise/security deployment, and continue core patent prosecution and runtime architecture development.

Yes. A private live walkthrough of Pontex in action is available for prospective investors and strategic partners. Submit the form or contact partnerships@pontex.ai to request access.

Because this is a Rule 506(c) offering, accredited investor status must be independently verified before an investment is accepted. Pontex uses a third-party verification process so sensitive financial verification documents do not need to be submitted directly to Pontex.

Pontex was founded by David Evans, Co-Founder & CEO; Justin Del Giudice, Co-Founder & COO/CFO; and Richard Santos, Co-Founder & Technical Inventor.

Potential international investors may be eligible subject to applicable securities laws, jurisdictional requirements and investor verification. Please contact the Pontex team before investing.