Featured
Live
DELOS AG
First paying enterprise customers in sixteen weeks
Highway · Freight-tech · Platform engineeringLive engagement
$180K a year off the AWS bill — first savings live in 27 days
Highway's product was scaling fast. The AWS estate under it was still run by hand. Streaver made it declarative — and took six figures off the annual bill inside the first month.
- Client
- Highway
- Sector
- Freight-tech · Logistics
- Engagement
- Embedded platform team
- Duration
- 6 months · ongoing
- First of its kind
- Highway's first outside engineering partner
~$180KAnnualized AWS savings
27 daysKickoff to first savings live
22MDB writes/day we operate in
2M+Loads the platform is built for
Highway · product preview
See how Highway works
A short look at the carrier-identity platform our embedded team keeps declarative, observable, and cheap to run.
Highway · product preview
01The Client
The carrier identity layer for freight.
- Company
- Highway
- HQ
- Dallas, TX
- Category
- Carrier identity · freight intelligence
- Scale
- 22M writes/day
Highway builds the carrier identity layer for freight — the system brokers and shippers rely on to know who they're actually doing business with. It's infrastructure for a fast-moving industry, and it has to behave like it.
By the time Streaver came in, Highway was scaling hard: tables taking tens of millions of writes a day, some of the largest RDS instances on AWS, and a platform built for well past two million loads. The product was working. The infrastructure under it was still being run by hand.
02The Challenge
A fast product on hand-run infrastructure.
Highway didn't have a reliability problem so much as a manageability one. The estate had grown faster than the discipline around it — and the gaps showed up where they hurt most: in releases and on the bill.
01
Run by hand
ECS services, scaling rules, and capacity were managed manually — with no single source of truth for what was actually deployed, or why.
Infrastructure02
Configuration drift
Image tags, tagging conventions, and sidecars had drifted between services and environments — small inconsistencies that quietly accumulated risk.
Drift03
Drift blocked releases
Inconsistent logging, tagging, and sidecar setup had become release blockers. The infrastructure itself was slowing the team down.
Delivery04
Sized by intuition
CPU and memory were provisioned from experience and caution, not production signal — so the estate was over-provisioned in places and fragile in others.
Capacity03Why Streaver
A trusted extension, not external help.
This was the first time Highway had trusted an outside engineering partner with its infrastructure. The engagement had to earn that on day one — and prove, fast, that it was a trusted extension of the team rather than external help.
Value in the first meeting
On day one — January 2 — the team spotted the EC2 → Savings Plans opportunity that became the first chunk of savings. Delivery-first, before any roadmap.
A trusted extension, not a vendor
As the months went on, Highway's leadership pulled Streaver deeper — relying on the team's Terraform and AWS judgment rather than treating it as staff augmentation.
Owners, not ticket-takers
Calls were made with the business in mind: what to standardize, what to leave alone, and what to flag for discussion rather than silently change.
04The Drop
From sprawl to a single source of truth.
The estate didn't need a rebuild. It needed a source of truth — so every change could be reviewed, traced, and justified with data.
Manual ECS configsEnvironment driftTribal knowledgeHand-tuned capacityInconsistent taggingUnapplied changes
One Terraform source of truth
Every service, secret, and scaling rule — declared, reviewed, and promoted through CI/CD.
05The Solution
Declarative, and legible by design.
Streaver embedded a platform team and worked in two directions at once: make the estate declarative, and make its behavior legible. Five workstreams carried the engagement.
01 · Declarative infrastructure
ECS under Terraform
The Monitor and Sidekiq services moved to Terraform — the first real source of truth for what's deployed and how to promote it. CI/CD gained safer plan / promote / apply flows, deliberate task-definition handling, and no more invisible unapplied changes.
02 · Secrets & config
A model, not a pile
A base secret per environment with optional per-service extension: common values live in one place, exceptions are explicit and controlled, every change is auditable, and duplication drops.
03 · Capacity
Evidence-based, not intuition-based
CPU and RAM were re-allocated from real production metrics — CPU, memory, queue pressure, Sidekiq latency, task counts — each change carrying a confidence level and rolled out in batches: high-confidence and low-risk first, with conservative thresholds for critical services.
04 · Observability
Signals you can act on
Puma backlog metrics, CloudWatch and Grafana, and New Relic / Grafana Alloy / Beyla sidecars surface real service pressure — and feed autoscaling, so the platform responds to load instead of to guesses.
05 · Operational judgment
Knowing what not to automate
Drift from incident-time manual fixes is preserved, not overwritten. Capacity providers are classified by cost and resilience. And the Sidekiq recombination question was handled as analysis — isolation, criticality, ownership, autoscaling — not a change shipped for its own sake.
06The Results
Cheaper to run. Safer to change.
Six months in, the estate is cheaper to run, safer to change, and able to justify every technical decision with data. Figures with a TBD badge are awaiting Highway's sign-off before we quote them publicly.
Manually managed~$180K/yr saved
AWS spend
Savings Plans on database and compute, plus right-sized capacity — the first cut spotted in the very first meeting.
Quarters, typically27 days
First savings
From kickoff on January 2 to the first savings live on January 29.
Manual & driftingDeclarative
Infrastructure state
ECS under Terraform as the source of truth; drift eliminated as a release blocker.
Out of syncSynchronized
Environments
Dev, stage, and prod brought into alignment so changes promote predictably.
IntuitionEvidence + confidence
Capacity decisions
Reallocations driven by production metrics, each with a confidence level and a staged rollout.
Hand-managed22M writes/day
Scale operated
The platform now runs declaratively at Highway's current scale, built toward 2M+ loads.
07The Stack
AWS, made declarative.
Cloud
- AWS
- ECSFargate + EC2 capacity providers
- RDSHigh-write Postgres
Infrastructure as code
- Terraform
- Shared modulesFrom official AWS modules
- CI/CDPlan / promote / apply
Observability
- CloudWatch
- Grafana
- New Relic
- Alloy / BeylaSidecar telemetry
Runtime
- SidekiqBackground jobs
- PumaBacklog-aware autoscaling
Who this is for
Scale-ups running a serious AWS estate by hand — where infrastructure inconsistency has started to block releases, and capacity is still a matter of intuition.
08The Team
The embedded platform pod.
Nico
Platform / DevOps EngineerInfrastructure as code and the CI/CD plan / promote / apply flows.
Chacal
Platform / DevOps EngineerEvidence-based capacity, the secrets model, and observability.
Tino
Platform / DevOps EngineerAWS cost optimization, RDS, and autoscaling signals.
Six months, traced
Jan 2, 2026KickoffDay one. The EC2 → Savings Plans opportunity is spotted in the first meeting.
Jan 29, 2026First savings live27 days in, the first cost reductions are running in production.
Months 1–6The estate goes declarativeECS under Terraform, a real secrets model, evidence-based capacity, and observability feeding autoscaling.
NowA trusted extensionOngoing. Streaver is pulled deeper into the platform as the partnership compounds.
Infra slowing your releases and inflating your bill?
Let's make it declarative.
Streaver embeds platform teams into scale-ups running serious AWS estates by hand — turning drift and intuition into a declarative, evidence-governed platform that's cheaper to run and safer to change.
