Mode A — Project by project
Every brief starts at zero.
- New foundations every time
- Knowledge locked in one repo
- Cost scales linearly with work
- Nothing compounds
Engineering team · Istanbul
Cities of engineering to empires of vision.
We build AI into real products — our own models, our own apps, and the infrastructure underneath them.
- Team
- 11 engineers, designers & media
- Units
- Eclipse · Nebula · Aurora
- Base
- BTM Kamp, Fulya · İstanbul
- In production
- ÇTH Portal
01 Positioning
We don’t ship projects.
We ship systems that keep shipping.
Most software work starts from zero every time. Ours doesn’t.
Mode B — Reusable systems
The foundation is already running.
- One layer under many products
- Models and agents we own
- Each build speeds up the next
- Capability compounds
Rebuilt every time Written once, reused
One team.
Eleven people. Flat. The founder is still in the repo.
Multiple systems.
Infrastructure, model pipelines and agent layers, reused across every build.
Multiple products.
A team set up to start the next one without starting over.
02 What we do
Three arms.
One engineering core.
The same people and the same infrastructure, pointed three ways.
01
Product Engineering
Our own AI models, mobile apps and web products.
Model trainingFine-tuningLoRA
Agent architecturesRAGMCP
On-device AI
02
Automation
The manual loop, replaced by a system that runs itself.
n8n workflowsAI agentsMail automation
Process automationSystem integration
03
Software Engineering
Mobile, web, backend — engineered as systems, not deliverables.
Mobile appsWebsitesBackend systems
Internal platformsCustom software
03 How we build
Depth over surface.
What we build, what we refuse to outsource, and what we keep.
P01
Reusable infrastructure
Auth, data access, agent runtimes, deploy pipelines. Written once, carried forward.
P02
Our own models
We build and fine-tune rather than routing everything through someone else’s service.
P03
Fine-tuning & LoRA
Adapters shape a base model to a real requirement. Narrow, testable, cheap to iterate.
P04
On-device AI
Inference close to the user. Lower latency, lower cost, data that stays put.
P05
Agent architectures
Tool calling, ReAct loops and MCP — agents that do work, with the failure modes written down.
P06
Scale through systems
The output isn’t a repo. It’s a loop that makes the next build cheaper.
Reference architecture
Request path & reuse loop
Scroll the diagram sideways
04 In development
Agent layer
ReAct runtime
SkyEngineX Mobile
An Android product with the agent layer built in — tool calling and a ReAct loop on a clean, testable architecture.
PlatformKotlin · Jetpack Compose
ArchitectureClean Architecture · MVVM
StackHilt · Room
Agent layerTool calling · ReAct
Planned launchOctober 2026
ModelFree to use · Enterprise licensing
In active development — the screen shows a representative agent trace, not a shipped interface.
05 Selected work
Systems that are actually running.
One live platform, one product in development, one research track.
Live in production
ÇTH Portal
A production portal coordinating employees, administrators and customers for a cleaning services company. An n8n-built AI agent supports the customer request flow, handling on the order of ten requests a day.
In development
SkyEngineX Mobile
Kotlin and Jetpack Compose on Clean Architecture and MVVM, with an internal agent layer using tool calling and ReAct. October 2026, free with enterprise licensing.
Research
Vision-language research
An ongoing research track in vision-language models, evaluated doctor-in-the-loop.
Research only — not a released product. No medical validation, no regulatory approval, and no claim about accuracy.
06 Engineering stack
The parts we actually run.
The layers a SkyEngineX system is assembled from.
Layer 01
Mobile
- Kotlin
- Jetpack Compose
- MVVM
- Hilt
- Room
- Clean Architecture
- React Native
- Expo Router
- Cross-platform mobile
- Frontend development
- Component-based architecture
Layer 02
Backend
- Java
- Spring Boot
- Ktor
- Spring Data JPA
- Hibernate
- Spring Security
- JWT authentication
- OAuth2
- RESTful API development
- MVC architecture
- Microservices
- SQL
- PostgreSQL
- MySQL
- Supabase
Layer 03
Web
- HTML / CSS
- TypeScript
- Node.js
- React
- Next.js
- Tailwind
- Three.js
- GSAP
- Vercel
Layer 04
AI/ML
- Python
- Machine learning
- Deep learning
- Supervised learning
- Unsupervised learning
- NumPy
- Pandas
- Scikit-learn
- PyTorch
- TensorFlow
- Keras
- JAX
- LLM fine-tuning
- LoRA
- RAG
- Vector databases
- pgvector
- MCP
- Agent architectures
Layer 05
DevOps
- Docker
- Containerization
- Cloud native
- CI/CD pipelines
- Git / GitHub
- Linux
- Server architecture
- Infrastructure management
- Deployment processes
- Backend deployment
- API integrations
- Monitoring & logging
Layer 06
Automation
- n8n
- Mail integrations
- Workflow orchestration
- System automation
- Agent-integrated automation
07 Team & culture
Eleven people, no hierarchy,
everyone in the code.
Build young. Think long-term. Ship fast.
Student-founded and student-driven. That isn’t a caveat — it is why we move quickly and keep the structure flat.
11People
8Engineers
3Design & social media
Unit 01
Eclipse
Mobile and AI products. Turning a model into something you can hold.
Unit 02
Nebula
R&D and systems engineering. The parts everything else stands on.
Unit 03
Aurora
Design and social media. How the work looks and how it travels.
Hüseyin Cem SallanFounder & Chief AI Engineer
Ömer ÇayırCo-Founder & Head of Technical
• Where we stand
Anthropic Claude for StartupsAccepted into the program.
BTM Kamp, FulyaWorking from the BTM Kamp office space in İstanbul.
Live customer platformA production system serving a real business today.
Team certifications
AnthropicNVIDIAIBM
Hugging FaceDeepLearning.AI / Stanford
08 Join the team
We’re looking for builders.
Not a CV drop. Tell us what you built and what you want to get good at.
- A real project in your first weeks, not shadowing one.
- You work directly with the people who started this.
- We care what you’ve shipped, not which year you’re in.
- Applications are open to people aged 18 and over.
09 Contact
Tell us what you’re building.
hello@skyenginex.com
Cities of engineering to empires of vision.
hello@skyenginex.com SkyEngineX · İstanbul