01 · Infra
Cloud infrastructure & platform
Architecture, migration, and day-2 operations across GCP, AWS, and Azure. Kubernetes, Terraform, GitOps, and the guardrails that keep cloud bills and blast radii small.
Software engineering company
We build systems
that hold up
in production.
A senior engineering practice. We design, build, and operate cloud infrastructure, data platforms, and AI systems for clients, and we ship our own software products. Small team, no hand-offs, no junior bench.
Services
Engineering, not staffing.
We take ownership of outcomes: an environment that deploys cleanly, a pipeline that doesn't page you at 2am, a system your team can actually operate after we leave.
02 · Data
Data & backend systems
Databases, APIs, and pipelines built for the load you actually have. PostgreSQL, Redis, event streams, and the migrations nobody else wants to touch.
03 · AI
Applied AI systems
LLM-powered features and agentic workflows integrated into real products, with evals, observability, and cost control. Not demos that die in a notebook.
04 · Reliability
DevOps & reliability
CI/CD, monitoring, alerting, and incident response. We instrument what we build and hand you dashboards that tell the truth.
05 · Product
Product engineering
Full-stack delivery from spec to shipped: React/TypeScript front ends, Python and Node back ends, and the release process in between.
06 · Leadership
Fractional engineering leadership
Architecture reviews, technical due diligence, and hands-on direction for teams that need a senior voice without a full-time hire.
Work
Workflows we've shipped.
A few representative engagements, with client details anonymized where needed. Every number below is measured against real billing, traffic, or telemetry, not estimated.
Cloud cost
Cutting a legacy cloud bill by a third
A 2001-era platform on GCP with no trended view of spend, and one cost line that had grown 16,128% quarter-over-quarter unnoticed. We ran a 14-day audit before touching anything, then re-platformed licensing and right-sized compute.
36%
run-rate cut, measured against billing actuals
Read the breakdown →
AI systems
4–7× throughput in a pipeline stuck at 15% CPU
A forecasting pipeline throttled to 75 items/hour with 85% of paid LLM capacity idle. The bottleneck was not capacity: it was phase-locked request waves. Diagnosed with numbers, fixed in one reversible PR.
4–7×
throughput unlocked, zero new infrastructure
Read the breakdown →
Product build
CurlyFry: a production AI SaaS, end to end
Our own engineering-management platform: React 19, FastAPI, Kubernetes with GitOps, live Stripe billing, and an LLM cost-governance gateway that keeps AI features economically survivable.
659
commits from empty repo to operated SaaS, in ~3 months
Read the breakdown →
Security audit
The password that sat in production for 7 years
Full-codebase reviews of legacy platforms with custom scanning rules and human triage. We rank findings against live traffic, so the fix list is ordered by real exposure, not severity labels.
479
confirmed findings across 11 repos; the worst fix was one line
Read the breakdown →
Approach
How we work.
01
Senior people only
Every engagement is staffed by senior engineers who have run production systems. You talk to the people doing the work.
02
Own the outcome
We scope to results, not hours. If the deploy pipeline is the problem, the engagement ends when deploys are boring.
03
Leave it operable
Documentation, runbooks, monitoring, and handover are part of the work, not an add-on. Your team should not need us forever.
04
Dogfood everything
The tooling and practices we recommend are the ones we use to run our own products. If it doesn't survive our stack, we won't sell it to you.
Contact
Tell us what's broken.
Or what you want to build.
One message starts it. We'll reply with questions, not a sales deck, usually within a business day.
✓
Got it. Thanks.
Your message is in. We'll reply with questions, not a sales deck, usually within a business day.