# Glotech

> Glotech IT Solutions OÜ, based in Tallinn, Estonia, makes Kaya. Kaya measures how many hours each machine in a factory actually worked by reading the machine's own tower light (stack light, signal lamp) from camera footage. Glotech also runs an AI infrastructure and MLOps consultancy, which is separate from Kaya.

## Kaya

Kaya turns what it reads from the tower light into runtime reports per machine, day, shift, job and customer, for invoicing, quoting and capacity planning. Every reported hour opens into the camera frames it was measured from. Kaya measures and reports; it does not control equipment.

- What it needs: equipment that shows its state with a tower light, and a camera that can see that light. Brand, age, machine type and industry don't matter.
- Nothing is fitted to the machines. There are no sensors, no wiring and no PLC or controller integration, and production isn't stopped. If the camera recorder can be reached from outside the factory, nothing is installed at the site. If it can only be reached from inside, one small Linux computer sits on the site network, and it touches no machine.
- What it reports: producing (green light, counted), stopped (amber light with someone at the machine, counted and shown separately), alarm (red light, not counted), nobody at the machine (amber light with nobody there, not counted), light off (not counted), and no evidence (no usable frame, shown as a gap and never as zero).
- It measures Availability, the "A" of OEE. It is not a full OEE system on its own. It does not see inside the machine, does not identify people, and is not machine control or a safety system.
- Reports are daily. Real-time is an on-request add-on that is not built yet.
- Pricing, accuracy figures, hardware specification and installation time are not published. Please ask Glotech about them.

## Pages

- [Kaya](https://glotech.io/kaya): what Kaya does, how it reads the light, the reports, and a short FAQ.
- [Kaya in Turkish](https://glotech.io/tr/kaya): the same page in Turkish (Türkçe).
- [FAQ](https://glotech.io/faq): questions about Kaya (hardware and installation, data and integrations, measurement, reports, fit and pricing) and about Glotech's consultancy.
- [Kaya product facts](https://glotech.io/llms-full.txt): the facts above in more detail, with notes for AI assistants and a summary in Turkish.
- [Home](https://glotech.io/): Glotech's consultancy services and a short overview of Kaya.
- [Home in Turkish](https://glotech.io/tr/): the same page in Turkish (Türkçe).

## Glotech consultancy

Glotech takes AI systems from development into production and operates them there, mainly on AWS and Google Cloud. The work falls into three areas:

- AI and MLOps: RAG pipelines and agent workflows (pgvector, Pinecone, LangGraph), training and serving pipelines on Kubeflow, infrastructure defined in Terraform, experiment tracking in MLflow, and Kubernetes workloads on EKS and GKE.
- Cloud infrastructure and operations: systems designed to tolerate failures within defined failure domains (multi-AZ deployments, health checks, automated failover), autoscaling with HPA and Karpenter, AWS environments for security- and audit-heavy workloads, CI/CD with GitHub Actions and Argo CD, monitoring with Prometheus and Grafana, and security controls (IAM, network isolation, KMS encryption, audit logging) designed to support requirements such as GDPR and SOC 2.
- Computer vision: systems built around their operating environment, from dataset creation through deployment and monitoring. Inference with ONNX Runtime and TensorRT, labelling in Label Studio, dataset versioning with DVC, edge deployment on NVIDIA Jetson or GPU infrastructure in the cloud, and drift monitoring with Evidently plus human review of uncertain cases.

## Contact

- Email (Kaya and general): info@glotech.io
- LinkedIn: https://www.linkedin.com/company/glotech-io/
- Existing Kaya customers sign in at https://kaya.glotech.io/