Inventory of Assets (TRL 4 Evaluation)
Frontend-ready assets valued for Kazakhstan market and global (UK/US) equivalents.
| Project Asset | Description | Market Value (QAZ) | Global Value (UK/US) |
|---|---|---|---|
| soile.dqaz.ai | Speech AI / Audio Capture Interface | $1,000 | $4,000 |
| hrtech.dqaz.ai | HR-Tech / Data Collection Forms | $1,200 | $4,500 |
| biz.dqaz.ai | B2B UI / Analytics Dashboards | $1,500 | $5,000 |
| edtech.dqaz.ai | EdTech / Offline Simulation Platform | $2,000 | $6,000 |
| Total | $— | $— | |
Astana Hub Nomination Strategy (TRL 5 Path)
Four pilot narratives aligned to measurable outcomes and deployability.
- Interview simulation + scoring rubric
- Language-aware evaluation (KZ/RU)
- Audit-ready decision trails
- Voice verification for step completion
- Offline manuals cache on edge
- Field-ready high-contrast UI
- Chat + voice playback interface
- Edge routing for low-latency responses
- Offline-first fallback flows
- AI multi-translator for hospitality
- Infrastructure health telemetry widget
- Tenant isolation and policy enforcement
The Pitch & Defense Kit
3-minute script + Q&A defenses for panel scrutiny.
- KZ markets need offline + bilingual AI that works under real constraints.
- We deliver TRL 4 assets now, and a clear TRL 5 pilot path.
- Four product lines: HR, EdTech, Travel, Biz gateway.
- Unified architecture: edge routing + policy + audit logs.
- Show live demos (SPAs) + pilot KPIs (latency, cost, compliance).
- Ask: Astana Hub nomination support + pilot partnerships.
Security: How do you secure corporate data?Expand
Answer: Zero-Trust by default: tenant isolation, policy enforcement, and auditable request paths.
- Edge-first routing with strict access controls.
- Per-tenant keys, scoped tokens, and audit logs.
- Data minimization with configurable retention.
Language localization: How do you handle KZ/RU bilingualism reliably?Expand
Answer: Localization is a first-class feature: UI text, transcripts, and evaluation rubrics are bilingual by design.
- Language detection + explicit controls per flow.
- Domain glossaries with QA loops.
- Terminology consistency checks.
Offline capabilities: What works without connectivity?Expand
Answer: Offline mode is supported via local edge nodes with cached safety manuals and limited inference paths.
- ESP32-S3-assisted offline content + verification flows.
- Store-and-forward sync when online.
- Operational continuity for critical training.
Scalability: Can it serve many users and requests?Expand
Answer: Designed for high-throughput edge routing and horizontally scalable inference and storage.
- Target capacity: 10,000+ requests per minute.
- Autoscaling API/translation/scoring behind stable endpoints.
- Separation of UI/API/inference/storage/audit.
Competition: How is this different from ChatGPT and generic tools?Expand
Answer: Not a general chatbot: controlled corporate workflows with offline pathways, auditability, and domain-specific UX.
- Workflow-first (HR, safety, hospitality routing).
- Enterprise controls (policy, logs, compliance posture).
- Latency targets via edge routing + low-latency inference layers.
Technical Data Sheet
Specification summary for TRL 5 readiness.
- NVIDIA NIM
- Vertex AI
- Cloudflare AI Gateway
- 10,000+ requests per minute (design target)
- Edge routing for low latency
- Horizontal scale across services
- 100% Personal Data Protection (Local Edge Nodes)
- Tenant isolation + policy controls
- Auditable request trails