Shape how LLMs understand the world.

Ensure coordination within the infrastructure team and enable efficiency across all engineering.

• Design and maintain deployment pipelines, cloud infrastructure, and monitoring
• Set standards for reliability and incident response
• Ensure smooth team growth and culture transfer to new employees
• Execute on the infrastructure strategy — tiered deployments, reliability, observability

Stack: Nix/NixOS, Terragrunt/Terraform, GCP (BigQuery), Airflow DAGs, Postgres, OTel/Grafana

Must have:

• Deep GCP or equivalent cloud experience in production
• Nix/NixOS fluency — we use it everywhere
• Terraform/Terragrunt — and strong opinions on making it more reliable or replacing it
• You've led infra for a team, not just yourself

Nice: Rust, multi-region design, Terragrunt alternatives

Own internal platform tooling — Airflow, model maintenance pipelines, deployment systems.

• Own and maintain Airflow pipelines, new model onboarding, and production deployment systems
• Ship cross-stack features across existing systems with high code quality
• Light SRE — upgrades, metrics, alerting improvements
• No on-call

Stack: Rust, Go, Python · Airflow DAGs · GCP (BigQuery) · Postgres · Nix/NixOS · OTel/Grafana

Must have:

• Production Rust or Go; Nix basics
• GCP basics; comfortable jumping into Python when needed
• You've maintained production pipeline orchestration (Airflow or equivalent)

Nice: OpenTelemetry, Grafana/Prometheus, Terragrunt

Close loops: accurate docs, reliable demos, small tweaks that make promises true.

• Build demo presets, safe defaults, and API endpoint tweaks
• Manage long-running system state and cross-service integration
• Write docs that people actually read

Stack: Rust/Go/Python · GCP (BigQuery) · Airflow DAGs · Postgres · Nix/NixOS · Terragrunt · OTel/Grafana

Must have:

• Solid engineering in one language; cross-stack curiosity
• CI comfort; knack for tidiness and developer experience

Nice: Nix, Terragrunt, metrics/logging, SQL

Build the agentic tooling that takes every employee — not just engineers — to 100× of their unassisted output.

• Own the developer loop — local dev, CI, the editor-to-deploy path. If `git push` makes anyone nervous, that's your bug to fix
• Build agentic tooling for the whole company — sourcing pipelines, analyst copilots, ops automations — wired into real systems, not toy demos
• Write the evals that tell us whether an agent works *every* Tuesday, not just the one we shipped on
• Maintain MCP servers and shared agent infrastructure so every tool we ship gets first-class access to our data
• Be the loudest voice in the room when an agentic approach won't work — the goal is 100× of real output, not 100× of plausible output

Stack: TypeScript, Rust, Python · Claude Agent SDK, MCP · Nix/NixOS · GCP · evals and orchestration

Must have:

• Production experience building agentic tools — code agents, LLM-driven workflows, internal copilots — not just prompt tinkering
• Strong systems engineering in at least one of Rust, Go, TypeScript, or Python
• You think evals before features; you've felt the difference between an agent that worked once and one that works reliably
• Comfortable writing tools for non-engineers without condescending to them and without dumbing the mechanics down

Nice: MCP server authoring, Claude Agent SDK or equivalent, Nix, prior internal-platform or tooling-team experience, PLT background

Make sure every team can deploy and use what engineering builds — without hand-holding from the infra team.

• Own developer experience across the company — define what "easy to deploy" actually means and make it real
• Build deployment guides and internal tooling for non-engineers
• Help business, sales, and design teams adopt infrastructure outputs
• If someone can't wrap their head around a deployment tool, the deployment strategy isn't good enough — fix the strategy, not the person
• Work in close collaboration with the Lead QA Engineer — your jobs inform each other

Stack: Nix/NixOS, GCP, macOS, Python, Rust/Go · internal tooling

Must have:

• You've helped non-technical people use technical systems — and didn't resent it
• Strong written communication
• Comfortable with multiple languages and environments including macOS

Nice: Nix, GCP, macOS systems administration, workshop/training facilitation, internal tools experience

Design and run the training and eval pipelines that maintain our lead in understanding AI marketing.

• Continue-pretrain and fine-tune models at scale (7B–35B+)
• Work with data filtering and formatting pipelines
• Run ablations and rigorous evals; maintain reproducible experiment infrastructure

Stack: PyTorch, FSDP/ZeRO · vLLM · Hardware: In-house H200 supercomputer; even larger H100 clusters for full-weight LLM training with Adam

Must have:

• End-to-end training experience (7B–35B+)
• Distributed training know-how (FSDP/ZeRO)
• Clean experiment hygiene — you log everything and your runs reproduce

Nice: Rust orchestration, retrieval-free evals, dataset curation instincts

Train and evaluate the models that power our core product.

• Run training jobs at scale and debug GPU-level issues when they surface
• Build and maintain evaluation infrastructure
• Experiment with training recipes — data mixes, hyperparameters, schedules

Stack: PyTorch, FSDP/ZeRO · vLLM · Hardware: In-house H200 supercomputer; even larger H100 clusters for full-weight LLM training with Adam

Must have:

• Experience training language models (fine-tuning at minimum, pre-training preferred)
• PyTorch fluency
• Comfort with distributed training concepts

Nice: vLLM, FSDP/ZeRO hands-on, data pipeline work, GPU cluster scheduling

Build the analytics infrastructure that turns raw data into business-relevant insights.

• Design and maintain data models and transformation pipelines
• Build reliable, tested data transformations that business teams depend on
• Ensure data quality, consistency, and documentation across the analytics stack
• Collaborate with analysts and engineering to define metrics and KPIs

Stack: SQL, Python · dbt · GCP (BigQuery) · Postgres · data modeling

Must have:

• Strong SQL — you think in data models, not just queries
• Experience with dbt or equivalent transformation tooling
• BigQuery or similar warehouse experience
• You care about data quality and documentation as much as feature delivery

Nice: Python, Postgres, data governance, semantic layers, data visualization

Extract and communicate insights that drive product and business decisions.

• Own business intelligence — define, track, and communicate key metrics
• Conduct deep-dive analyses that inform product and go-to-market decisions
• Build dashboards and reports that non-technical stakeholders actually use
• Collaborate with engineering on data model requirements and quality

Stack: SQL, Python · BigQuery · data visualization (Metabase, Looker, or similar) · statistical analysis

Must have:

• Strong SQL and analytical thinking
• Experience communicating data insights to non-technical stakeholders
• You've influenced product or business decisions with data — not just produced reports
• Statistical literacy — you know when a trend is signal and when it's noise

Nice: Python, BigQuery, A/B testing, product analytics, data storytelling

Architect enterprise-grade network topologies for a company that will change how traffic is served at scale.

• Architect enterprise-grade, complex network topologies for full-stack traffic shaping
• Lead reverse proxy infrastructure — design and ship production-grade reverse proxies
• Own and evolve the in-house traffic classifier
• Set technical direction for the network engineering team as it grows
• Cross-collaborate with ML on sophisticated bot analysis and detection models

Stack: Rust or Go · Caddy (advanced) · GCP MIG autoscaling · HTTP/TLS/DNS · iptables/nftables · in-house traffic classifier

Must have:

• You've designed and operated enterprise-grade network topologies in production
• Deep TCP/IP stack expertise — sockets, routing, datacenter-level traffic flows
• Experience with reverse proxy architecture at scale (Caddy, Nginx, HAProxy, or equivalent)
• Comfort with iptables, nftables, UFW, or equivalent packet filtering

Nice: OpenWrt, Caddy module development, GCP managed instance groups, security engineering, Rust or Go

Build full-stack traffic shaping infrastructure for a company competing at Cloudflare's level of ambition.

• Build and maintain network infrastructure for full-stack traffic shaping with GCP MIG autoscaling
• Develop and extend our in-house traffic classifier for bot detection and fingerprinting
• Push Caddy to its limits — advanced reverse proxy configurations, dynamic upstreams, custom modules
• Contribute to differentiated traffic serving — different responses for scrapers versus humans
• Cross-collaborate with ML on detection model analysis

Stack: Rust or Go · Caddy (advanced) · GCP MIG autoscaling · HTTP/TLS/DNS · iptables/nftables · in-house traffic classifier

Must have:

• Deep understanding of TCP/IP, HTTP, TLS, DNS, and how data flows across datacenters
• You know your sockets, routers, and packet filtering (iptables, nftables, UFW)
• Experience with web crawling, scraping, or traffic shaping at scale

Nice: OpenWrt, Caddy module development, GCP managed instance groups, Rust or Go, protocol analysis

Own code quality, fault tolerance, and reliability across all engineering teams.

• Design, implement, improve, and maintain testing strategies across the stack (unit, integration, E2E)
• Build CI quality gates that catch real bugs, not just lint noise
• Continuously maintain our monolith deployment tools — ensure zero drift and always-workable condition
• Establish testing culture and mentor engineers on testing practices
• Work in close collaboration with the Developer Experience Engineer — your jobs inform each other

Stack: Rust, Python, Go · CI/CD · Vitest/Playwright · property-based testing · Nix

Must have:

• Strong testing methodology — you know when to unit test and when to integration test
• Experience with CI/CD quality gates in a real shipping team
• You've caught bugs before they became incidents
• Enough seniority to set quality bars and push back when needed

Nice: Rust testing patterns, property-based testing, chaos engineering, Nix, monolith deployment

Solve the dichotomy of performance versus safety — achieve speed within our deliberately safe architecture.

• Profile and optimize critical paths within our safety-first architecture
• Design database schemas and queries that are both safe and performant — we use patterns like Postgres event sourcing deliberately
• Build benchmarks and performance regression tests
• Advise teams on achieving performance within architectural safety constraints, not by removing them

Stack: Postgres (event sourcing) · Rust, Python · profiling/benchmarking · BigQuery

Must have:

• Deep Postgres or equivalent RDBMS experience
• You understand that removing guardrails is not performance engineering — you find speed within the architecture of safety
• You've profiled and fixed real performance problems in production
• Comfortable with Rust and/or Python

Nice: Event sourcing, functional programming, BigQuery optimization, distributed systems performance, load testing