The Role
You are the architect and the engineer. You design a system on paper, then build it yourself, own it in production, and set the standard the rest of the team works toward. You will own one full workstream end-to-end: design → implementation → testing → deployment → operations.
What We Need
- Distributed systems design Seven or more years of engineering, with distributed systems you designed and built at production scale — and can describe precisely. You define the problem when nobody else has, document decisions as ADRs, and settle arguments with benchmarks rather than opinions.
- Expert Python Pydantic models, asyncio, dataclasses, typing throughout. You understand the GIL and when it actually matters.
- Data pipelines Dagster orchestration, medallion architecture, event-driven design.
- Declarative configuration Native fluency with YAML and configuration-driven design.
- Data engineering Arrow and Parquet, partitioning, schema versioning, entity resolution, deduplication, data lineage.
- Designing for failure Retry budgets, circuit breaking, graceful degradation, replay semantics, no single points of failure.
- Operations Kubernetes, Docker, Helm, CI/CD discipline, OpenTelemetry and structured logging.
- Technical leadership You mentor senior engineers on system design, name technical debt, and tell product the difference between infeasible and over-engineered.
Tech Stack
- Language: Python, Pydantic, asyncio, dataclasses, typing.
- Orchestration: Dagster, Ray, event-driven design.
- Data: Arrow, Parquet, MinIO, Elasticsearch, Redis, PostgreSQL, Neo4j, Qdrant, Weaviate, Milvus.
- AI Infrastructure: vLLM, Ollama, embedding services, H200 SXM cluster.
- Platform: Kubernetes, Docker, Helm, GitLab CI/CD, GitHub Actions.
- Security: Air-gapped deployment, Keycloak, OIDC, mTLS, RBAC/ABAC, HSM/KMS, zero trust.
- Observability: OpenTelemetry, Prometheus, Grafana, Tempo, data lineage.
Strong Advantage
- Knowledge graph design (Neo4j, TigerGraph) or vector databases.
- LLM deployment on-prem, air-gapped, or in sovereign environments.
- Intelligence, security, or investigative data platforms.
Not a Fit
- Architects who don't code, or coders who don't architect.
- Engineers who wait for tickets.
- Anyone who would ask Claude what architecture to use instead of knowing.
What We Offer
- Full ownership of a critical system component, not a queue of tickets.
- Small team, direct access to the CEO, no bureaucracy.
- Remote-first, with structured monthly in-person collaboration.
- Compensation for genuine senior talent rather than market rate for the title.