Renkler tuhaf mı görünüyor? Samsung Internet tarayıcısı koyu modda site renklerini değiştiriyor olabilir. Kapatmak için Internet menüsünden Ayarlar → Kullanışlı Özellikler → Labs → Web site koyu temasını kullan seçeneğini etkinleştirebilirsiniz.
Corporate technical training · RuneLab Academy
Run your AI on your own systems.
Hands-on training on running open source LLMs on your organization's servers. From model selection to GPU planning, from serving with vLLM to on-premise RAG and a chatbot interface, your team builds the whole system itself; data never leaves your network.
Either can be the right answer. The decision depends on how sensitive the data is, the volume of use and your team's operational capacity. The training teaches the criteria for making it.
For technical teams that will stand up and run open source models in-house. Familiarity with Python, Linux and Docker is enough.
Mid-level and senior developers who want to learn the open source LLM ecosystem and build AI applications that run in-house.
Teams that will plan GPU capacity and build and scale the model serving layer.
Operations specialists who will own deployment, monitoring and maintenance of local LLM systems.
Leaders choosing between cloud and on-premise based on cost, performance and security.
In local LLM projects the first question is hardware. Pick a model size and a precision to see how much GPU memory the model weights take. In the training you do this calculation for your own workload, including context length and the number of concurrent users.
Want this hardware installed and ready? Discover RuneBox
Seven modules. Each one is hands-on in a live GPU environment; examples are adapted to your organization's use cases.
The Qwen, Mistral, Gemma, gpt-oss, DeepSeek and GLM families; choosing a model by task, size, Turkish performance and license. GGUF and SafeTensors formats, reading a model card.
Estimating VRAM; how AWQ, GPTQ, FP8 and GGUF quantization affect memory and quality. Single server or cluster, tensor and pipeline parallelism across GPUs, containers with Docker.
PagedAttention and continuous batching; connecting existing applications unchanged through an OpenAI-compatible API. Quick prototypes with Ollama and llama.cpp, production with vLLM and SGLang.
Workshop Publishing a model in-house as an OpenAI-compatible endpoint.
Multilingual embedding models such as BGE-M3 and multilingual-e5; setting up Qdrant, Milvus and pgvector on-premise and choosing between them for your use case.
Document processing for PDF, DOCX, XLSX and email, OCR for scanned documents, chunking, retrieval and reranking. Managing models and packages in air-gapped environments, network segmentation and encryption.
Workshop An end-to-end RAG system with every component running in-house.
Interfaces with Open WebUI, Chainlit and Gradio; conversation history, source citations and feedback. Authentication with SSO, integration with the intranet.
Measuring TTFT and tokens/s, load testing; optimization with batching, the KV cache and semantic caching. Monitoring with Prometheus metrics, API keys and rate limits. Total cost of ownership, cloud vs on-premise.
Output A cloud vs on-premise cost comparison for your organization.
The weight of each module, the models used and the exercises are agreed in a scoping call before the training, based on your team's profile and your existing infrastructure.
At your office; the exercises are prepared around your infrastructure and use cases.
Live sessions for teams in different locations; can be split into half-days.
Conceptual modules online, workshop modules on-site; for large groups.
Every participant receives a RuneLab Academy certificate of attendance. On request, the training is run together with Boğaziçi University Lifelong Learning Center and the certificate is issued with BÜYEM endorsement.
Each layer is built in its own module. By the end, your team has built the whole system once, end to end, and keeps the setup steps, configurations and code.
Corporate network
Module 7 · Performance and monitoring
The ready-made version of the system you build in the training: it comes installed on your premises with NVIDIA-based GPUs, local models and RAG infrastructure, and RuneLab maintains it.
Explore RuneBox Used together with RuneBox: RuneDesk · RuneGate
A local LLM is a language model that runs on the organization's own servers instead of a cloud API. The model is loaded onto in-house GPUs; questions, documents and answers never leave the corporate network. This training teaches, hands-on, how to build and run such a system with open source models.
The length is set in a scoping call before the training, based on the scope of the modules, your team's level and whether the exercises run on your own infrastructure, and it is stated in writing in the quote.
No. The GPU environment for the exercises is provided during the training; how participants reach it is planned with you beforehand, around your network restrictions. If your organization has GPU infrastructure, part of the exercises can also run on it.
No. Exercises use sample document sets we prepare; no real company data is entered into any tool.
Open model families such as Qwen, Mistral, Gemma, gpt-oss, DeepSeek and GLM. The choice depends on the task, Turkish performance, hardware and license terms. Model versions are reviewed before every training.
Yes. Once the models, embedding models and vector database are installed, the system runs without any outside connection. Moving models and packages into an air-gapped environment, updates and access control are covered separately in the program.
This program focuses on running existing open source models on-premise. Fine-tuning with LoRA and QLoRA is planned separately when needed.
The RAG Training focuses on designing systems that answer from corporate knowledge with cited sources, and it applies to any model, in the cloud or on-premise. This training teaches running the model itself on-premise; here RAG is built as a fully on-premise pipeline. Explore the RAG Training.
RuneBox is the architecture taught in this training, delivered installed and maintained by RuneLab. If you want your own team to build and run the system, choose the training; if you want to start with ready-made infrastructure, choose RuneBox. Teams using RuneBox also take this training to run the system and build applications on it.
Every participant receives a RuneLab Academy certificate of attendance. On request, the training can be run together with Boğaziçi University Lifelong Learning Center and the certificate issued with BÜYEM endorsement.
The training is delivered by experienced senior AI engineers and data scientists who run enterprise AI projects at RuneLab.
Yes. In a short call before the training we learn about your team's profile, your existing infrastructure and your use cases, and prepare the models and exercises accordingly.
We prepare a quote for your organization based on the number of participants, the format and how much the content is tailored. Fill in the form and we will get back to you within one business day.
Tell us the number of participants, your team's technical profile, and your preferred format and dates; we will get back to you within one business day.
