Company Overview
IFORELS Inc. (brand: iFrame®) is a Menlo Park, California–based technology company founded in 2022 by Vlad Panin. While the 8-figure ARR business is run by a solo founder, the company operates through official agreements with universities worldwide and partnerships with Tier 1 cloud providers.
iFrame® has evolved through three distinct stages — from an R&D lab focused on solving fundamental GPU limitations with a custom GPU Operating System, through a healthcare AI application phase that reached 995K users, to its current revenue-focused business model: AI-Infrastructure-as-a-Service.
IFORELS Inc.’s core technical innovation is a proprietary approach to overcoming GPU data range limitations, using NVMe storage as a buffer for GPU memory and managing decentralized calculations on remote GPUs via iFrame® OS. This breakthrough forms the foundation for a virtualization technology that dramatically reduces GPU compute costs at 4–6× below prevailing market rates.
Key Facts
| Legal Entity | IFORELS Inc. |
| Brand | iFrame® |
| Founded | 2022 |
| HQ | Menlo Park, CA |
| Founder | Vlad Panin (sole founder) |
| Team | Founder + 4 PhD contractors |
| Domains | iframe.ai |
| Focus | GPU-as-a-Service / Inference-as-a-Service |
| Key Tech | Custom GPU OS · Infinite context inference · NVMe-buffered GPU memory |
| Users (peak) | 990K (*at Medical Coding AI) |
Links
iframe.ai
med.report
sefirot.ai
pulsar.global
panin.one
Stage 1: R&D Lab — Infinite Context & GPU Memory Innovation
IFORELS Inc. began as a research-focused lab working on a fundamental problem in AI infrastructure: the limitations imposed by GPU data range constraints (such as FP32 precision range) and the finite size of GPU memory (VRAM), which cap the context window size of large language models.
The team’s approach was to use NVMe storage as a buffer for GPU memory — effectively extending the working memory available to AI models beyond the physical VRAM of the GPU. This required solving complex challenges in data transfer orchestration between NVMe and GPU at the OS level, maintaining computational precision across the extended memory hierarchy, and ensuring that inference speed remained practical despite the additional data movement.
The result was a technology that enables effectively unlimited context windows — allowing AI models to process inputs far exceeding the limits of conventional transformer architectures. The company filed a patent related to this technology, which was confirmed registered by late 2025.
This R&D work produced two named models — Asperanto and Sefirot-10 — based on an underlying architecture called “Monoidal Framework.” Inference was published on a separate domain, sefirot.ai, with the ability to search data across the internet in real time. In August 2025, the company publicly announced the technology, describing it as the world’s first “Large Attention Model” (LAM) with an infinite context window that had been tested on inputs exceeding one billion tokens.
Stage 2: Healthcare AI — Medical Coding (990K Users)
After successfully testing the infinite context window technology, IFORELS Inc. focused on the healthcare market, starting with services billing administration, specifically medical coding. The iFrame® product was a proprietary AI model providing automated medical coding for ICD-9, ICD-10, ICD-11, HCPCS, and CPT codes, including all modifiers, integrated with EHR systems via the HL7 protocol.
The platform featured 186 boilerplate AI agents organized by clinical category. Users could copy, fine-tune (by uploading insurance terms, regulations, clinical specifics), and redistribute custom agents (“iFrames”). Additional capabilities included automated insurance preauthorization via phone, fax, and email, document generation by template, and AI-powered voice interaction.
Growth
iFrame® reached 995K users within approximately 18 months with a near-zero marketing budget. Growth was driven entirely by useful free content — practical medical coding tools and educational resources — that attracted organic search traffic. At its peak, the platform was generating tens of thousands of daily impressions from search engines and hundreds of new user registrations per day at zero customer acquisition cost.
Go-to-Market Challenges
Despite strong user adoption, the bet on large hospitals purchasing the product did not work due to the hypermonopolized EHR market in the United States. A small number of dominant EHR vendors control access to hospital systems, creating significant barriers for new AI tools seeking enterprise integration. Individual healthcare practitioners and billing specialists adopted iFrame® enthusiastically, but converting this bottom-up adoption into enterprise-level purchases proved extremely difficult within the existing market structure.
This experience led directly to the Q3 2025 pivot: rather than continuing to compete for access within a monopolized market, IFORELS Inc. redirected its core technology toward a market with fewer structural barriers and faster sales cycles.
Stage 3: AI-Infrastructure-as-a-Service (Q3 2025 – Present)
In Q3 2025, IFORELS Inc. pivoted to AI-Infrastructure-as-a-Service, leveraging the unique virtualization technologies and custom OS-level orchestration of data transfer and calculation developed during the R&D phase.
The value proposition is straightforward: IFORELS Inc.’s GPU operating system makes existing cloud GPU infrastructure dramatically cheaper for end users. The iFrame® OS technology reduces the cost per hour of current-generation NVIDIA GPUs by 4–6× compared to prevailing market rates, by optimizing how data flows between storage, memory, and compute at the operating system level.
Rather than owning GPUs or data centers, IFORELS Inc. operates as an infrastructure optimization layer — partnering with neocloud providers and their clients to deliver more cost-efficient AI training and inference environments.
Revenue Trajectory
The GPU-as-a-Service business showed rapid acceleration from its first signed contract in early 2026:
Timeline | Milestone |
|---|---|
| First contract | Initial neocloud client signed on a 6-month term. Revenue expected within days. |
| ~Week 2 | Multiple long-term contracts (1–2 year terms) signed. |
| ~Week 3 | Continued growth. Data center expansion planning initiated. |
| ~Week 4 | Acceleration in signed contracts continues. |
| ~Week 5 | Significant contract portfolio established, excluding a pending large multi-year deal in pipeline. |
| ~Week 5–6 | Additional 12-month contract signed. |
In addition to signed contracts, the company has a pipeline deal for a full NVIDIA DGX SuperPOD on a 5-year commitment. Equipment has been specified, colocation with confirmed power capacity has been secured, and the client has indicated readiness for a significant prepayment. Target launch is mid-2026.
Core Technology
IFORELS Inc.’s iFrame® OS offering is built on three technical pillars developed during the R&D phase:
Custom GPU Operating System: A proprietary OS-level layer that orchestrates data transfer and calculation across GPU, NVMe, and system memory. This eliminates bottlenecks that cause conventional GPU cloud setups to underutilize hardware, recovering significant performance typically lost to virtualization overhead.
NVMe-Buffered GPU Memory: By using NVMe storage as a high-speed buffer for GPU VRAM, the system overcomes the physical memory limitations of individual GPUs. This enables larger models to run on fewer GPUs, directly reducing cost per inference or training job.
Infinite Context Inference: The same unlimited context window technology that powered the healthcare AI product is now available as an infrastructure capability. Clients can run models with context sizes far exceeding conventional limits, opening use cases in enterprise document processing, large-codebase analysis, and multi-modal reasoning that are impractical on standard GPU infrastructure.
Infrastructure
IFORELS Inc. currently operates on major cloud provider infrastructure and is expanding toward dedicated hardware. The planned 72-server data center expansion (NVIDIA DGX SuperPOD) has secured colocation with 2.5MW of confirmed power capacity and equipment specified through a major hardware distributor. Technical support operations are approximately 99% powered by AI, enabling the five-person team to maintain enterprise-grade service levels.
Market Context
IFORELS Inc. is entering the rapidly growing GPU cloud market at a time of significant structural change. GPU cloud pricing has dropped substantially from 2023 peaks, and the market spans several tiers: major hyperscalers with broad ecosystems at premium pricing, well-funded neocloud specialists, mid-tier GPU clouds, and emerging decentralized compute networks.
IFORELS Inc.’s differentiation — a 4–6× cost reduction via custom OS-level virtualization and orchestration — positions the company not as a direct competitor to GPU cloud providers but as an optimization layer that makes their infrastructure dramatically more cost-effective. The rapid acceleration from first contract to a significant portfolio of signed agreements within approximately five weeks suggests strong product-market fit in a segment where cost optimization is a primary purchasing criterion.
How It Started
Vlad Panin personally invested $200,000 to start IFORELS Inc. in 2022, funding the initial R&D lab and the early development of the infinite context window technology. Between 2023 and 2025, our amazing angel investors contributed a combined $225,000 in SAFE. As the company grew through its healthcare AI phase and into the GPU pivot, another $270,000 from personal savings throughout 2025 to keep the project going — covering remote team costs, A/B tests, and marketing experiments during a period when the company was transitioning between business models and had not yet secured meaningful revenue.
Funding Summary
Source | Amount | Purpose |
|---|---|---|
| SAFE | $695,000 | Product development, Marketing |
| Credits Google | $900,000 | Computational Resources |
| Credits AWS | $5,300,000 | Computational Resources, R&D team support |
| Total | $6,895,000 |