Example portfolio · NoirCreate yours free→

Priya Raghunathan

Staff engineer — platform & reliability

I work on the layer nobody sees until it breaks. Mostly distributed systems, capacity, and the slow institutional work of making on-call survivable. I have deleted more services than I have launched, and I consider that the accomplishment.

Bengaluru, India

Email
4Experience
4Projects
16Skills

Experience

Staff Engineer, Platform

KettleBengaluru2022 — Present
  • Technical lead for the compute platform running roughly 900 services.
  • Drove p99 request latency down 41% by replacing the service mesh sidecar path.
  • Cut cloud spend 23% in one year through bin-packing and honest capacity reservations.
  • Rewrote the on-call rotation and runbook model; pages per engineer per week fell from 9 to 2.

Senior Site Reliability Engineer

KettleBengaluru2020 — 2022
  • Owned the incident process end to end during a period when the company tripled in size.
  • Built the error-budget policy that finally gave product teams a shared language for reliability tradeoffs.

Infrastructure Engineer

Alloy SystemsPune2017 — 2020
  • Migrated a monolith's data layer from a single Postgres primary to a sharded topology, online, over seven months.
  • On-call for the storage tier — the education you cannot get any other way.

Software Engineer

Vantis TechnologiesChennai2015 — 2017
Backend services for a payments product. Learned what an idempotency key is the expensive way.

Projects

Sidecar removal

Kettle platform · 2023
Replaced per-pod proxy sidecars with a node-local data plane across 900 services. p99 latency down 41%, roughly 12% of total cluster CPU returned. Migration ran service-by-service behind a flag for four months.
  • #Distributed systems
  • #Kubernetes
  • #Performance

Error budget policy

Kettle · 2021
An org-wide SLO framework with real teeth: budget exhaustion freezes feature deploys for that service. Adopted by 40 teams. The interesting work was social, not technical.
  • #SRE
  • #Process

Online resharding

Alloy Systems · 2019
Moved a 14TB Postgres primary to a 16-shard topology with no maintenance window, using logical replication and a routing layer that could fail back at any point.
  • #Postgres
  • #Migration

shelf

Open source
A small Go library for bounded in-process caches with correct TTL semantics under contention. About 2,400 stars; used in a few places I did not expect.
  • #Go
  • #Open source
github.com/praghunathan/shelf

Education

National Institute of Technology, Trichy

B.Tech Computer Science & Engineering2011 — 2015

Skills

Languages

Go, Rust, Python, SQL, Bash

Infrastructure

Kubernetes, Terraform, Envoy, Kafka, Postgres, Redis, ClickHouse

Observability

Prometheus, OpenTelemetry, Grafana, distributed tracing

References

Priya is the person we send into the systems nobody wants to own. She comes back with a plan that is smaller than you expected and works.
Daniel OkaforVP Engineering, Kettle
The error budget policy changed how this company argues about reliability. That was not a technical achievement, and she knew it going in.
Sana MirzaPrincipal Engineer, Kettle

Talks

Deleting the Sidecar

SREcon Asia-Pacific2024
On what we gave up to get 41% back, and why it was worth it.

Error Budgets Nobody Ignores

KubeCon India2022
A policy is only real if it can stop a launch.