From hardware to model, model to application.All in one stack.
Our own GPU cloud, our own models, our own platform. We run the same engine in our products and in your operations — KVKK-compliant, Turkish-optimized, on-prem ready.
AI ürünlerini geliştir,
entegre et ve yayınla
Agent'lar oluşturun, servisleri API olarak kullanın ve çözümlerinizi hızla yayına alın.
Agent Oluştur
Chatbot Yayınla
API ile Entegre
$ curl -X POST https://api.siriusai.com
-H "Authorization: Bearer YOUR_API_KEY"
-d { "model": "sirius-llm-70b", ... }
Four layers, one owner.
Hardware, model, platform and application — end-to-end ours. The integration burden of multiple vendors is not delegated to you. Use any layer on its own or take the full stack.
Application
State-machine agent flows via Flow Designer. Calling AI, Seferi and Boss AI run on this layer. Built-in template gallery: Municipal, Telecom, Finance, Healthcare.
Platform
AI Studio: 10 modules, OpenAPI-compliant REST endpoints, JS/TS and Python SDK, visual UI, webhook and MCP support. One key, every service.
Model
Turkish-native LLM (8B, 70B), prosody-aware TTS, sectoral STT, embedding and image models — all trained from scratch by our team. Fine-tuneable on customer data.
Hardware
GPU cloud inside Turkish borders. TensorRT-LLM compilation, custom CUDA kernels, KV-cache and speculative decoding — markedly lower latency and cost than open-source inference servers.
Ten modules, one API key.
Every module is production-ready and consumable on its own. Test in the web UI in minutes, integrate via REST API right away.
LLM Playground
Turkish-native 8B and 70B models. TensorRT-LLM compilation, KV-cache and speculative decoding for low-latency response.
Metin
Merhaba, randevunuzu yarın saat 10:00 için onaylıyorum.
Text → Speech
Neural voices trained from scratch for Turkish prosody. First-byte latency in the hundreds of milliseconds — sufficient for live voice agents.
Speech → Text
Dialect-aware, tuned for finance/health/telecom terminology. Word-level streaming output.
Mazeret izni yıl içinde toplam 5 iş gününü geçemez.
RAG
Hybrid BM25 + vector retriever, source-grounded answers that minimize hallucination. Documents stay in an isolated vector DB.
Fatura.pdf
{
"tarih": "2026-05-04",
"tutar": "₺ 4.870,00",
"vkn": "6840547852",
"kdv": "%20"
}
OCR
Turkish ID, driver's license, invoices, policies. Template-based parsing into structured JSON.
Flow Designer
Combine LLM, RAG, function and decision blocks into a single state machine. The backbone of Calling AI.
Image Generation
Brand-tuned generation. Production runs in an isolated GPU queue. Capacity opens on request.
Moderation
Trained on data labelled for Turkish slang, hate speech and sectoral risk. Thresholds calibrated to your policy.
Turn Detection
Diarization optimized for call-center scenarios. Low-latency streaming support.
Mesafe
128 km
Süre
2s 45dk
Yakıt
−18%
Logistics Optimizer
Powered by the Seferi engine. Solves with time windows, capacity and driver break constraints.
Sectoral, production-deployed flows.
Examples of scenarios we built with our customers and shipped to production. Each one is a single state machine running in Flow Designer.
Call Center
01 / 06Voice Customer Service Agent
Transcribes the inbound call in real time with SiriusAI STT, detects intent, fetches customer data via RAG, generates a resolution and replies naturally with TTS. Escalation rules live in Flow.
Municipal / Public
02 / 06Complaint & Routing Line
Branches emergencies to 112 and standard requests to a ticket flow; auto-fills location and category slots. Turkish dialect awareness drives high accuracy.
Telecom
03 / 06Tech Support + Billing Flow
Manages modem-reset steps, billing queries and routing branches in a single flow. Connects to customer records via RAG; hands unresolved calls to a human operator.
Finance
04 / 06Card & Fraud Trigger Flow
At trigger points like card block, charge dispute or lost/stolen reporting it authenticates the user, fires KYC steps and makes system calls through function tools.
Logistics
05 / 06Route & Load Optimization Assistant
Interprets driver voice updates, recomputes the route via Logistics Optimizer and reports the new ETA to the customer with TTS.
Healthcare
06 / 06Appointment & Triage Assistant
Takes the patient's complaint in natural language, routes to the right department via RAG and reserves a slot through the appointment system. Sensitive data never leaves the institution.
One engine, five products, in production every day.
Calling AI's real-time voice agent engine, Seferi's route and fleet optimizer, Boss AI's document-grounded CEO copilot, Lobster Lead's content engine and Talentius's HR ecosystem — all run on the same AI Studio layer. The platform you will integrate is the very same engine our products rely on every day.
Products Running on AI Studio
One platform, four ways to consume it.
Pick by speed and technical capacity. Move up a tier as your needs grow.
Self-Service Platform
Start instantly via the web UI and API. Open an account, generate a key, make your first call within minutes. Ideal for SMEs, product teams and rapid POCs.
Ready-Made Products
Plug deployed agents like Calling AI, Seferi and Boss AI directly into your operation. The modelling burden is not on you.
Managed Service
We set it up, we run it. Our ops team rolls it out with 24/7 monitoring, model versioning and an observability dashboard.
Custom Development
Greenfield modelling, custom flow design and application. A vertical-specific agent family built end-to-end by our team.
You decide where the data runs.
Our GPU cloud inside Turkey, a federated hybrid architecture, or your own on-prem infrastructure.
SiriusAI Cloud
Our TR GPU cloud
Lowest cost, fastest start. Best for SMEs, product teams and rapid POCs.
Hybrid
Data on you, model on us
Federated/proxy architecture. For organizations sensitive to data leakage but unwilling to take on the install.
On-Premises
All on your infrastructure
Finance, public sector, healthcare. Install, license and support package included.
From off-the-shelf to a custom model, in four levels.
Start fast, deepen as your needs grow. Each level builds on the previous one.
- L1
Configuration
HoursParameters, system prompt, voice selection, flow parameters. Self-service, no extra fee.
- L2
Fine-tuning
1–2 weeksTuning the base model with customer data. Deployment under a custom name.
- L3
From-Scratch Training
4–12 weeksFully bespoke model production (LLM/TTS/embedding). All IP with the customer, isolated GPU.
- L4
Flow + Application
2–6 weeksEnd-to-end scenario + UI built on Flow Designer. Within the custom solution scope.
Enterprise oversight, sectoral compliance.
Processing within Turkish borders, auditable records, certified infrastructure.
Data Sovereignty
All SiriusAI Cloud resources sit inside Turkish borders. KVKK-compliant processing, user rights and deletion guarantees are standard.
Encryption
AES-256 at rest, TLS 1.3 in transit. API keys are managed with rotation, scope limits and rate limiting.
Audit & RBAC
Every API call, flow step and model call is logged. Role-based access, team management, principle of least privilege.
Sectoral Compliance
ISO/27001 certification in progress, BDDK compliance documentation for finance, annual independent audit report for the Enterprise tier.
Let's plan your AI transformation together.
In a 30-minute discovery call we'll choose the right package, modules and deployment model for you.