Glotech.io
AI Architecture
We build RAG pipelines on pgvector and Pinecone, and design agent orchestration with LangGraph. Guardrails keep retrieval grounded, so every output traces back to its source documents.
We build AI systems that keep
running in production
Technologies we use
Kubeflow
Google
Cloud
Terraform
Prometheus
Grafana
Docker
Python
Kubernetes
AWSCore Capabilities
MLOps Engineering
We deploy training and serving pipelines on Kubeflow, provision infrastructure with Terraform, and track experiments with MLflow. Any environment can be rebuilt exactly, and a change gets from commit to production faster.
Platform Operations
We run Kubernetes clusters on EKS and GKE, and automate node scaling and cost allocation with Terraform. Prometheus and Grafana keep infrastructure observable and costs in check.
What We Do
High Availability Systems
We configure multi-AZ deployments, automated failover through Route 53 health checks, and self-healing Kubernetes pods. Recovery from an outage stays under a minute.
Auto-Scalable Infrastructure
We provision Kubernetes HPA and Karpenter-based node autoscaling on AWS and GCP. Clusters scale with real load and right-size themselves to keep cloud spend in check.
Cloud-Native AWS Architectures
SageMaker for model training, Lambda for event-driven logic, S3 for data lakes, CloudWatch for alerting: we combine these into AWS architectures built for workloads with strict compliance and audit requirements.
DevOps & Observability
We set up CI/CD with GitHub Actions and ArgoCD, monitor services with Prometheus and Grafana, and configure PagerDuty alerting. Teams ship daily with full visibility into what's running.
AI System Integration
We connect ML models to production applications through typed API contracts, structured logging, and circuit breakers. Predictions are served behind clear failure boundaries, with fallback paths in place.
Security & Compliance Engineering
We implement IAM policies, VPC isolation, encryption at rest with KMS, and audit logging for SOC 2 and GDPR. We put these controls into the deployment pipeline on the first day, so nobody has to add them later.
Computer Vision
We build computer vision systems that run in production, from the first dataset to deployment. We design them to be reliable, easy to improve, and suited to the conditions they will actually work in.
Production-Ready Vision Systems
We build inference pipelines with ONNX Runtime and TensorRT, served behind FastAPI endpoints with health checks and structured logging. Every prediction is traceable.
Data-Centric Development
We manage datasets with DVC for versioning and Label Studio for annotation. In our experience, careful review of the data improves accuracy more than moving to a bigger model.
Training, deployment and retraining
We train RT-DETR and custom detection models with reproducible MLflow experiments, deploy through canary rollouts, and retrain on feedback from production.
Edge & Cloud Deployment
We deploy vision models to edge devices (Jetson, Coral) for low-latency inference, and to GPU-backed Kubernetes for cloud workloads. Latency, throughput, and cost get profiled before anything goes live.
Dataset Management & Versioning
We version datasets with Git LFS and S3-backed storage, so training runs can be compared side by side. When a model regresses, the issue traces back to the exact data change that caused it.
Monitoring, Drift & Human-in-the-Loop
We monitor prediction confidence with Prometheus, flag distribution drift with Evidently, and route low-confidence samples to human reviewers. Models stay accurate as real-world data shifts.
Kaya
Kaya reads the tower light on each machine through the cameras your factory already has, and reports how many hours each machine actually ran. Nothing is fitted to the machines.
Reading the tower light
Kaya reads each machine's tower light as green, amber, red or off, and turns that into hours per machine, per day and per shift. When the light is amber and nobody is at the machine, that time is shown separately and not counted.
Nothing fitted to the machines
Kaya uses the cameras already on the factory floor. You don't add sensors, wiring or a PLC connection. If your camera recorder can only be reached from inside the factory, one small computer sits on your network, and it touches no machine. Brand and age don't matter: CNC, presses, injection moulding, packaging lines and any other machine with a tower light.
Every hour has its camera footage
Any hour in a report opens into the camera frames it was measured from, with the light colour Kaya read and how confident it was. If the camera could not see the light, that time is marked as no data, never as zero.
Frequently Asked Questions
What is Glotech?
Glotech is an AI infrastructure and MLOps consultancy focused on taking AI systems from development into reliable production.
Do you only work with AWS?
No. AWS is a major part of our work, but Glotech also works with Google Cloud and cloud-agnostic technologies such as Kubernetes, Terraform, Docker, Prometheus, Grafana, and MLflow.
Is Kaya a consultancy project?
No. Kaya is a product developed by Glotech for equipment-runtime measurement. Glotech's consultancy work and Kaya are separate offerings.