Elijah Shaw-Rutschman
Backend & Full-Stack Software Engineer · API design & integration, Auth (JWT/MFA/OAuth), AI platform tooling · Python, TypeScript, Nim
Summary
Software engineer with nearly two decades of experience across backend, platform, API, and full-stack work, now focused on AI developer tooling and platforms. At StyleSeat, a Python and Django marketplace, designed and integrated production APIs across the Stripe billing, subscriptions, and webhook stack, built a Stripe Radar fraud-detection system, and owned reliability and release tooling. Builds open-source AI tooling: spellbook, a multi-platform harness for AI coding assistants with a shared MCP server; momus, a provider-agnostic AI code-review GitHub Action; and soong, a GPU-instance CLI for AI agent workflows. Also built a Claude-powered code-review bot adopted org-wide at StyleSeat. Works across the stack with TypeScript and React. Earned StyleSeat's highest performance rating ("Exceptional") and was on the formal Staff Engineer promotion track. Chicago-based and open to remote roles across the US.
Skills
APIs & Auth
- REST API design
- API integration
- Webhooks
- JWT
- MFA
- OAuth integrations
- Access control
- OIDC
- SSO
- RBAC
AI & LLM Tooling
- AI-assisted engineering
- LLM application development
- AI agents
- MCP (Model Context Protocol)
- AI code review
- Claude Code
Languages
- Python
- TypeScript
- Nim
- Cython
- C/C++
- Rust
- SQL
Backend & Web
- Django
- Celery
- REST APIs
- React
- FastAPI
Data & Infrastructure
- MySQL
- Redis
- AWS
- Terraform
- Docker
- CircleCI
- GitHub Actions
- PostgreSQL
Payments
- Stripe billing
- Stripe subscriptions
- Stripe Radar
- Webhooks
- Fraud detection
Systems & Low-Level
- Lock-free concurrency
- Memory reclamation
- Realtime audio / DSP
- C / Cython interop
GPU & Compute
- GPU compute (Vulkan/SPIR-V)
- Compute shaders (GLSL)
- GPU memory management
- CUDA/Metal interop
Practices
- Release engineering
- Incident response
- TDD
- Mentoring
Work Experience(5)
Twelve years wearing many hats on a Python/Django marketplace: full-stack product, native mobile, search, payments, reliability, and most recently DevOps and platform engineering. Grew into a senior engineer the company relied on for its hardest, most cross-cutting problems.
Payments & API integration: Longtime key payments-platform engineer and the team's payments domain expert; major contributor to the Stripe billing and subscriptions build-out (promotion codes, plan and product backfills, coupon-stacking prevention, Stripe Portal, Pro Earnings tab), designing and integrating third-party APIs and webhook flows.
Authentication & API integrations: Built multi-channel MFA code delivery (SMS and email via Braze) and JWT authentication for the native-app Stripe billing-portal flow; worked across many API and auth integrations, including internal APIs with JWT, Braze, and Stripe webhooks.
AI code review: Built "StyleSeatBot," an org-wide automated code-review bot powered by Claude. It became a required reviewer on every pull request, signing off alongside a human approver, with adoption up to the CTO.
Reliability & failover: Owned a custom fault-tolerant Redis client library for graceful degradation when Redis was down, and designed a Celery enqueue-failover path that captured broker-outage failures as durable dead-letter records, plus an S3-backed idempotent recovery toolkit that replayed 351 tasks lost during a broker outage.
Rate limiting: Built StyleSeat's Redis-backed API rate limiter (per-IP and per-user, sliding window, kill-switch) with a DDoS-resistant local-memory cache layer to cut Redis load; applied across auth, payment, and registration endpoints.
Fraud detection: Architected and built an automated fraud-detection system on Stripe Radar.
Platform migrations: Led the pdm-to-uv migration; contributed to org-wide Python 2-to-3 (~900K LOC), Segment-to-Braze, and MySQL 5.7-to-8 migrations.
AI enablement: Helped drive the company's move to AI-assisted engineering as a member of the AI-adoption tiger team, running internal talks and workshops on Claude Code and LLM fundamentals.
Release engineering: Owned release tooling: a custom CircleCI orb, deploy-gate and hotfix/cherry-pick workflows, and build-failure alerting; contributed to CI/CD and Terraform/ECS.
Scale & mentorship: Among the highest PR-review and code-authoring volume on the team; delivered work across product, platform, payments, and infrastructure codebases, and mentored engineers.
Full-stack product engineering at a social-video startup, building Python/Django backends and JavaScript frontends across data models, APIs, and UI.
Led Django development of client web applications at a development studio, owning architecture and full-stack delivery across multiple engagements.
Built internal web applications and data tooling in Python/Django to support consulting engagements.
Developed web application features for a SaaS chemical-safety and compliance platform, moving into software engineering from an IT and internal-tooling role.
Projects(8)
- AI tooling
- MCP
- LLM
- Developer experience
Multi-platform harness for AI coding assistants: skills, hooks, and a shared MCP server across Claude Code, OpenCode, Codex, Gemini CLI, and ForgeCode. Anchor of the axiomantic org.
- Go
- MCP
- LSP
- AI tooling
Go MCP language server. Contributed a Streamable HTTP transport alongside stdio (stateful sessions and heartbeat), an LSP-to-MCP file-operations bridge, and ~12 additional LSP tools with capability-gated registration.
- Rust
- GPU
- Vulkan/SPIR-V
- Compute shaders
GPU compute in Rust over the Kompute Vulkan framework: a safe wrapper and C++ FFI bridge, plus hand-written GLSL compute shaders (tiled matmul, parallel reductions). Experimental.
- AI code review
- LLM
- GitHub Actions
- Python
AI code-review GitHub Action that works with any OpenAI-compatible LLM provider (Anthropic, OpenAI, OpenRouter, AWS Bedrock). Runs a four-phase pipeline (plan, review, verify, post) with sandboxed read-only repository tools and per-repo YAML configuration. Python and TypeScript.
- AI infrastructure
- GPU
- CLI
- Python
CLI for launching and managing Lambda Labs GPU instances for AI agent workflows: one-command launch with model selection, SSH tunneling with automatic port forwarding, and cost controls (lease limits and idle auto-shutdown). Python, MIT-licensed.
- Developer tooling
- libclang
- Cython
- C/C++
Cython .pxd binding generator for C and C++ headers. As the primary maintainer, rewrote it onto a libclang backend that handles real-world headers. 52+ stars, on PyPI.
- CI/CD
- QEMU
- Cross-architecture
- GitHub Actions
GitHub Action to run CI on non-x86 CPUs (ARM, aarch64, s390x, ppc64le, riscv64) via QEMU. Rewrote it to be more maintainable and testable, reviving community contribution. 747 stars.
- Nim
- Compiler internals
- Type systems
- Code generation
Merged fixes to the Nim compiler's type system and C code generator, including cycle detection for recursive concepts, sizeof for typedesc in generic template when clauses, C codegen for lifecycle hooks on generic distinct types, and object initialization for opaque importc fields.