Skip to content

Cloud, Data & Platform

85% infrastructure cost cut with a hybrid bare-metal platform

Managed databases, vector search and LLM API tokens had grown into a six-figure monthly bill. We moved the heavy paths onto owned infrastructure without giving up the cloud where it earns its keep.

Client
High-growth AI and analytics SaaS
Industry
AI & analytics SaaS
Disclosure
Anonymised under NDA
Lower recurring infrastructure spend
85%Lower recurring infrastructure spend
Query latency, down from 350 ms
<45 msQuery latency, down from 350 ms
Unencrypted network paths
0Unencrypted network paths

The challenge

Six-figure monthly cloud bills on managed databases, vector search and third-party LLM tokens, with analytical queries above 350 ms.

Stack

  • ClickHouse
  • Private GPU inference
  • WireGuard mesh
  • AWS

What we did

  1. 01Moved primary analytical storage to a self-hosted ClickHouse cluster
  2. 02Deployed private GPU nodes for local model inference
  3. 03Connected the owned estate back to the cloud over a zero-trust WireGuard mesh
Generic reference architecture
failoverTERRAFORMopentofu · helmPOLICY GATEopa · trivyLANDING ZONEiam · oidc · vpcREGION Aeks · auroraREGION Bwarm standbyOTEL · SLOgrafana · loki

A generic representation of the pattern, not the client’s topology.

How you would buy this

Embedded Architecture & Build

Near full-time senior leadership that also builds: platform shifts, private AI infrastructure and production incidents.

Compare all engagements →

Retainer · 30–40 hours a week

Embedded Architecture & Build

From

$15,000

per month

Near full-time senior leadership that also builds: platform shifts, private AI infrastructure and production incidents.

  • Hands-on engineering in your repositories
  • GPU, hypervisor and ClickHouse platforms
  • Private LLM gateways and agent pipelines
  • SLA-backed incident response
Ask about this engagement

More cloud, data & platform work

Cloud, Data & PlatformNDA · Financial services

Zero-trust multi-account landing zone that passed SOC 2 first time

Regulated financial services provider

A monolithic cloud estate with shared credentials and flat networking had an audit date. We rebuilt it as modular, policy-bounded accounts with every change in code.

Type II passed with zero findings on first evaluation
SOC 2Type II passed with zero findings on first evaluation
Infrastructure provisioned from code
100%Infrastructure provisioned from code
Public IP exposures left
0Public IP exposures left
Cloud, Data & PlatformNDA · Logistics

From kickoff to paying customers in 14 weeks

Series A logistics SaaS

A greenfield dispatch platform. We owned the architecture, the infrastructure and the first three service surfaces, then hired the team that took it over.

Kickoff to first paying customer
14 wksKickoff to first paying customer
Uptime from day one
99.95%Uptime from day one
Engineers hired and onboarded
3 → 11Engineers hired and onboarded
Cloud, Data & PlatformNDA · Healthcare

SOC 2 Type II and HIPAA readiness in one quarter

Pre-Series-A healthtech

A hospital pilot depended on SOC 2 and HIPAA controls the team did not have. We rewrote the infrastructure as code and built the evidence trail an auditor needs.

To SOC 2 Type II report
90 daysTo SOC 2 Type II report
Critical findings on first audit
0Critical findings on first audit
Hospital pilot closed
1Hospital pilot closed

Start with a scoping call.

Thirty minutes with the founder. Bring your architecture diagram, your cloud bill or your plant drawings, and leave with what is realistic — including when the answer is that you do not need us yet.

Mutual NDA before any detail is shared. First response within 24 hours.