OptimaPlatform is a Russian full-cycle ecosystem for building computer vision systems. We replace the departed Roboflow, Labelbox and Supervisely with a single domestic environment: from data preparation to model deployment and an ML-artifact marketplace.
of a CV project budget goes to manual data labeling.
Roboflow, Labelbox, Supervisely: sanctions, FX payments, data stored abroad.
sensitive data cannot be processed in foreign clouds.
popular OSS models under AGPL / non-commercial — a threat to commercialization.
Six stages in one platform + two cross-cutting layers — marketplace and reputation. 3–5× lower development effort.
Import, versioning, semi-automatic labeling with AI-assist (19+ models with verified licenses).
Training-code generation (LLM-Codegen) without manual coding; INT8 quantization and edge deployment.
Server and edge deployment, model-quality monitoring in production.
Platform control in natural language — a breakthrough UX for CV tools.
Selling datasets, weights, pipelines and services with escrow settlements and reputation. Network effect and retention.
An automated CI gate admits only Apache / MIT / BSD — removing legal risk for the customer.
Sources: Allied Market Research, Grand View Research, MarketsandMarkets; bottom-up estimate — 2000+ Russian AI teams × ARPU. The SOM projection reflects the commercialization plan and is not a guarantee of returns.
5 letters of intent from operating companies — a diversified B2B funnel, not “one customer, one case”.
Security systems integration
Reselling and co-marketing in the regional B2B market.
Building fire protection
Vertical solution for hazardous industrial facilities.
Central-station alarm monitoring
SaaS with intelligent video verification of events.
Edge/IoT device manufacturing
Quantized models and LLM-Codegen for mass-produced products.
Multispectral video analytics
R&D collaboration and licensing for CII facilities.
combined revenue forecast from partners by the end of year 3. Pilots in Q4 2026 → contracts in 2027.
| Parameter | OptimaPlatform | Roboflow (USA) | CVAT (OSS) | Yandex DataSphere |
|---|---|---|---|---|
| Full cycle “data → deploy” | ✓ | ~ | ✕ | ~ |
| Russian jurisdiction / on-premise | ✓ | ✕ | ~ | ✓ |
| Autonomous LLM control agent | ✓ | ✕ | ✕ | ✕ |
| ML marketplace with escrow | ✓ | ✕ | ✕ | ✕ |
| Model license-purity control | ✓ | ✕ | ✕ | ✕ |
| CII clearance / 152-FZ | ✓ | ✕ | ~ | ✓ |
| Entry price | from ₽0 (Free) | $ currency | ₽0 (labeling only) | ₽, generic |
Free / Pro ₽4,500 / Business ₽15,000 / Enterprise from ₽50,000 per month.
Enterprise and government: from ₽250,000/mo + license (air-gap, CII).
5–10% commission on escrow deals (after V2 launch). Education and white-label are additional streams.
| ARPU | ₽4,500/mo |
| Gross margin | ~75% |
| CAC | ~3 000 ₽ |
| LTV (24 mo) | ~81 000 ₽ |
| LTV / CAC | ~27× |
| Payback period | ~1 mo |
lines of proprietary code
ML models with verified licenses
in development since 31 Dec 2022
in operation, cloud pilot
Trained by the project team from scratch and shipped within the platform. It proves our R&D capability: we don’t just integrate third-party models, we build our own protectable neural-network solutions.
Intellectual property: applications filed for software registration (Rospatent) and the OPTIMAPLATFORM trademark. Demand confirmed by 5 letters of intent from customers in security systems, fire safety, alarm monitoring, Edge/IoT and video analytics.
R&D grants from Russian innovation-development institutions — non-dilutive and validated by expert review.
Venture capital to scale sales, marketing and grow the team.
SaaS and on-premise licenses generate operating cash flow from year one.
Open to dialogue with venture funds, angel investors and strategic partners. Detailed materials (business plan, financial model, pitch deck) on request.
Alexey N. Kiselev · PhD (Eng.) · founder and chief scientist of the project
This page is informational only and is not a public offer, individual investment advice, or a securities offering. The market estimates and projections are based on open sources and our own methodology and do not guarantee future results.