Reliability at scale (SRE)
Observability, SLOs, incident response, on-call, high availability. 30M requests and 100k jobs a day, a 500 GB database.
99.99%uptime, sole SRE
Platform Engineering & SRE
Production systems made faster, cheaper and more reliable, at scale.
Vincent Legendre, Senior Platform Engineer / SRE. Eight years in the field, formerly a data and software engineer. Sole SRE of a platform serving 30M requests a day. I scope, I ship, I hand over — on my own.
99.99 %Uptime, full yearSole SRE, 30M requests a day
10× lowerp99 latency
3× lowerInfrastructure billCompute moved to bare metal, egress eliminated
Method
How I work: pragmatism
The right solution, not the most impressive one.
I size the effort to the actual need
A controlled 30-minute cutover rather than a zero-downtime strategy, because the cost/benefit didn’t justify it.
I decide on numbers, not hunches
Profiling, EXPLAIN plans, response-time distribution analysis, all the way to the patch.
No vendor lock-in
Multi-provider architecture, with no dependency on a single supplier.
I ship things that last
Infrastructure as Code, tests and observability from the outset, documentation as a matter of course. The team carries on without me.
Expertise
What I do
Eight areas, from reliability to the data platform — each backed by a result, not a claim.
End-to-end web performance
Front end, back end and database treated as one system. Failed background jobs down by 95%.
2×faster API responses
Kubernetes, from migration to operations
Application migration (Helm, reproducible deployments, controlled rollbacks, zero-interruption cutover), picking up a half-finished migration, running it in production.
3production clusters
PostgreSQL in depth
Performance, high availability, replication, and the delicate operations done without downtime: version upgrades, pg_repack, Row Level Security with no regression.
0downtime on upgrades
Infrastructure cost (FinOps)
Cloud-to-bare-metal migration, egress eliminated, at least matching performance.
3×lower infra bill
Cloud architecture (AWS, GCP, Cloudflare)
Full architectures: authentication, compute, databases, IAM, API exposure, WAF. From the network to application security.
Industrialisation (CI/CD and IaC)
Terraform, Pulumi, Ansible; secrets and configuration under control. CI bill cut to a third.
<3 minCI, from 10+
Data engineering and platform
GCP-native data platform built end to end at a payments fintech (BigQuery, Dataflow / Apache Beam, self-hosted Elasticsearch, Airflow). Left the data team self-sufficient.
Tools
Stack
Orchestration & platform
- Kubernetes (k3s / Rancher)
- Helm
- CloudNativePG
- Cilium
- Flux CD
- Tailscale
IaC & CI/CD
- Terraform
- Pulumi
- Ansible
- GitHub Actions
Observability
- Datadog
Data & storage
- PostgreSQL
- Elasticsearch / ELK
- BigQuery
- dbt
- Metabase
Data engineering
- Apache Spark
- Apache Beam / Dataflow
- Airflow
- Prefect
Cloud & infrastructure
- AWS
- GCP
- Cloudflare
- OVH (bare-metal)
Languages
- Python
- SQL
- Scala
Engagement
How an engagement runs
Audit
A quick audit to size the work and put real numbers on the need.
Scoping
Scope and pace set once the work is framed — never announced blind.
Delivery
The deep understanding and development needed to ship. One change at a time.
Hand-over
I hand over: documentation and reproducibility. The team carries on without me.
Time and materials or fixed price, depending on how predictable the scope is. Fully remote.
Proof
Measured results
3× lower
Infrastructure bill cut to a third: compute moved to bare metal, egress eliminated.
At least matching performance.
10× lower
p99 latency brought down tenfold — while the average never moved.
99.99%
Uptime held for a full year, sole SRE, on a platform serving 30M requests a day.
100k jobs a day, a 500 GB database.
2× faster
API response times halved; failed background jobs down by 95%.
Through instrumented diagnosis, not rewrites.
<3 min
CI brought from over ten minutes to under three, CI bill cut to a third.
Unsure about a number in your production?
A scoping audit is enough to start: an objective diagnosis and a prioritised action plan, with no commitment beyond it.
Who it’s for
Who it’s for
SaaS scale-ups, fintechs, high-traffic platforms, growing engineering teams with no established SRE function. Typically a team of around twenty developers that needs senior Platform / SRE expertise without hiring full-time.
Request a scoping audit
An objective diagnosis and a prioritised action plan, with no commitment beyond it. Two fields are enough — the rest is quicker said out loud.
Book a slot
Thirty minutes, no commitment. Best if you already have a number that worries you.
See availability