Self-hosted LLM & MLOps in private AWS for UK HealthTech
Services Covered:

Industry
Software Development
Duration
2 months
Budget
10K - 49K
Start Date
1 December 2025
Team Size
1-5
Engagement Model
Offshore
Project Capability Score (PCS)
82/100
Strong
Project Capability Score (PCS) estimates how capable Honeycomb Software is of successfully delivering a project like this, based on its past experience, track record, and ability to handle similar work. Learn more about PCS
Client
Sonomics
Project Summary
Honeycomb Software implemented a secure, scalable AI infrastructure for a custom software engineering firm (Sonomics) to enable a HealthTech client’s self-hosted LLM within a private AWS environment. They delivered MLOps architecture, CI/CD/GitOps pipelines, GPU autoscaling (scale-to-zero), observability, and thorough runbooks to meet UK data residency and compliance requirements. The work produced a production-ready, compliant LLM capability with clear handover documentation.
Key Challenges
- Build a secure, scalable AI architecture the client can independently operate and expand
- Run an LLM entirely inside the client's private AWS environment
- Meet strict UK data residency and compliance requirements by keeping all AI processing and storage within the client's controlled infrastructure
- Minimize GPU infrastructure costs while ensuring production readiness and observability
Project Deliverables
- Kubernetes-based deployment of an open-source LLM in the client's private AWS environment
- MLOps architecture design and environment setup
- CI/CD pipelines and GitOps workflows for automated model deployments
- Kubernetes/GPU node management and scale-to-zero autoscaling policies
- Inference observability (metrics, logs, alerting) and operational runbooks
- Structured handover and documentation
Project Solution
Acting as a specialist subcontractor, Honeycomb designed and implemented a Kubernetes-based deployment of an open-source LLM within the client’s private AWS environment. They provided MLOps architecture and environment setup, CI/CD pipelines and GitOps workflows for repeatable deployments, Kubernetes/GPU node management with scale-to-zero autoscaling to cut idle GPU costs, and an inference observability stack with metrics, logs, and alerting. The engagement included operational runbooks and a structured handover so the client’s team could operate and extend the platform independently.
Project Outcome
- Compliant, production-ready self-hosted LLM running entirely within the client's UK AWS environment with no external API dependencies
- Reduced idle GPU costs via scale-to-zero autoscaling
- Full observability enabling confident production management of inference workloads
- Clean handover and documentation allowing the client to operate and extend the platform independently
- All deliverables were completed on time with clear communication and risk flagging
Platforms
Cloud
Tech Stack
Amazon Web Services (AWS)
Kubernetes
Client Endorsement
Overall Review Rating
5
5 out of 5 stars
Timeliness
Cost Rating
Willing to Refer
Quality of Deliverables
Honeycomb Software delivered the project, and now we have a compliant, production-ready self-hosted LLM capability. The team communicated clearly, owned their deliverables, and raised risks early. The quality of their handover documentation really stood out.
Note: This endorsement is based on publicly available client feedback from external review sources.
More Projects by Honeycomb Software
Endorsed
Completed
Information Technology
AWS Infrastructure & Frontend Development for Industrial DataLabs
Services Covered:
+4 more
Endorsed
Ongoing
Information Technology
Product Dev for Software Consulting Firm
Services Covered:
+1 more

