JOHN KELLY

[H]omeCV[D]ownload[T]heme

Staff software engineer specializing in global scale networks and distributed compute platforms. 8 years designing and operating production systems serving millions of requests per second and petabytes of data per day. spanning in-kernel packet processing, L4/L7 load balancing, bare-metal server automation, and cloud network architecture.

SKILLS

NETWORK
BGP, IPVS, Nginx, HAProxy, Envoy, CDNs, Internet & Web Protocols, Linux Kernel Networking, XDP/eBPF, WireGuard, VyOS
COMPUTE
KVM/Firecracker, Proxy-Wasm/WASI, Kubernetes, VMware, Proxmox, Ceph
INFRA
GCP, AWS, Postgres, Cassandra, Varnish, Packer, Terraform, Ansible, Saltstack
CODE
Rust, Python, Go, Java, TypeScript, C++

EXPERIENCE

Discord

San Francisco, CA · Remote

Lead Staff Software Engineer // Traffic Infrastructure

Dec 2025 — Present
Defined the network architecture and cross-cloud connectivity requirements for Discord's initial multi-cloud strategy, driving alignment across engineering, finance, and executive leadership
Built client-side network error reporting that surfaces regional and ISP-specific outages affecting users globally, giving policy and CX teams actionable path-quality signal that server-side telemetry cannot see
Designed an HMAC-signed header scheme for authenticating end-user requests at the CDN edge, enabling granular rate limiting and tiered abuse prevention based on client identity

Staff Software Engineer // Media Infrastructure

Feb 2024 — Dec 2025
Led multi-CDN initiative serving petabytes of static assets daily, reducing latency 15% and egress bandwidth costs 30% through measurement-driven provider steering
Developed internal content-addressable storage system enabling strong versioning of in-app assets uploaded to the global CDN, reducing developer friction and allowing immutable caching of highly fetched content at the edge.
Optimized the mobile attachment upload flow with more efficient image compression, achieving 30% faster upload latency

Senior Software Engineer // Traffic Infrastructure

Jul 2022 — Feb 2024
Migrated all inbound API traffic reaching ~2M RPS at peak to an Envoy-based proxy fleet, allowing for more precise traffic steering and circuit-breaking/bulkheading at the edge.
Drove service mesh adoption by migrating 90% of core production services across 12+ infrastructure and product teams, migrating all production services from legacy client-side routing to Envoy proxy sidecars
Designed and built a self-service traffic platform with TypeScript configuration, reducing new service infrastructure provisioning time 75%

Squarespace

New York, NY

Software Engineer // Traffic Infrastructure

Dec 2019 — Jul 2022
Built a custom DDoS detection and enforcement engine using XDP/eBPF, classifying and dropping attack traffic in-kernel at the earliest point in the datapath before it consumed userspace proxy capacity
Replaced legacy hardware L4 load balancers with a software IPVS solution using direct server return (DSR), improving tail latencies and high-availability failover across on-prem and cloud footprints
Reduced TLS handshake latency 15% by designing and building a new L7 proxy to offload TLS termination at the edge

Software Engineer // Compute & Storage

Jul 2018 — Dec 2019
Automated the hypervisor hardware lifecycle across a fleet spanning 20+ racks in two data centers, spanning iPXE/iDRAC bare-metal provisioning through Packer/Ansible image baking, with rebuilt images rolled out fleet-wide daily
Operated and capacity-planned 20+ Kubernetes clusters across on-prem data centers and cloud, alongside VMware virtual machine infrastructure and Ceph storage clusters
Built a Kubernetes operator to automatically provision and optimally schedule virtual machines, reducing developer toil and improving fleet resource efficiency
Cut base image build time 80% by reworking the Packer-based image pipeline

PROJECTS

AS401898 — Autonomous System & Multi-Tenant Network Fabric

Ongoing · Colocation, home, and cloud
Operate a production ASN with a direct ARIN IPv6 allocation (/40), running eBGP with an upstream transit provider from a colocated Proxmox/VyOS deployment
Designed a multi-tenant addressing plan with VRF-per-tenant isolation and WireGuard overlays spanning colocation, home, and cloud environments
Built a Python DSL that generates Ansible inventory from a declarative model of the fabric, making router and host provisioning repeatable across the estate

Sandboxed Execution — Firecracker / WASM Runtime

Draft — to be written
What the sandbox runs, and for whom.
The isolation boundary and why Firecracker or WASI was the right one.
A number: cold-start time, density per host, or throughput.

EDUCATION

Bachelor of Science, Computer Science

May 2018
Stevens Institute of Technology — Hoboken, NJ

CONTACT

LINKEDIN
GITHUB
EMAIL
j███@██████.ioCLICK TO REVEAL