Senior Full-Stack Engineer

The short version

We run an in-house recruitment tech platform that powers an entire business: candidate sourcing, BD pipeline, placements, predictions, MBR reporting, and a large set of AI-driven sourcing and enrichment workflows. It's mature, in production, and has evolved over roughly two years.

It's been built and maintained by one engineer. We're hiring the next one. You'd own it end-to-end (database, background jobs, AI pipelines, and frontend) and keep shipping.

This is not a green-field project, not a team role, and not a 9-to-5 ticket shop. It's real ownership of a real system that the business runs on.

What you'd be working on

A live, lived-in codebase, not a prototype. To give you an honest sense of the scale:

• ~415,000 lines of TypeScript / TSX across ~1,400 files
• 47 dashboard feature sections (153 page files)
• 90 API route groups across 383 endpoint files
• 164 Inngest background functions: crons, fan-out workflows, durable jobs
• 318 SQL migrations against an 84-table Supabase Postgres schema
• 250 React components, 377 library/util modules, 102 Vitest test files

You don't need to have all of this in your head on day one. You do need to be the kind of engineer who can drop into a large codebase, read it, and ship safely without hand-holding.

The stack

Frontend: Next.js 16 (App Router), React 19, Tailwind 4, shadcn/ui, with an in-house design system layered on top. Zustand, TanStack Table, Tiptap, Recharts.

Backend / data: Supabase (Postgres + Auth + SSR). Heavy RLS, custom RPCs, views, large junction tables, and 318 migrations of accumulated schema work.

Background work: Inngest, self-hosted via Docker on a DigitalOcean droplet (Postgres + Redis + Caddy alongside). 164 durable functions handling sourcing, scoring, enrichment, market tracking, deal-agent SMS, and call intel.

AI layer: Vercel AI SDK v6, Anthropic SDK (Claude Sonnet/Opus), OpenAI, Google GenAI, plus Mastra for agent/workflow orchestration (agents, tools, memory, evals, Langfuse tracing).

Workflow: Push-to-main. No PRs, no feature branches. We ship multiple times a day.

Where things stand right now

A few things are in flight, so you'd be stepping into live work, not a frozen system:

• A CRM design-system reskin, mid-migration
• An ongoing junction-table consolidation on the data model
• An active AI provider migration

Must-have skills

• 5+ years TypeScript / Node.js
• 3+ years React; 2+ years Next.js App Router (Server Components, Server Actions, route handlers)
• Strong Postgres, not just ORM-level. Real SQL, indexes, query plans, migrations, RLS
• Supabase, or equivalent Postgres + Auth + SSR client patterns
• Production LLM app experience: prompt engineering, structured output, tool calling, retries, cost/latency tradeoffs. Not "called the OpenAI API once"
• Durable background job systems: Inngest, Temporal, BullMQ, SQS or similar. Comfortable with fan-out, crons, idempotency, retries
• Tailwind plus a component library (shadcn/Radix, MUI, Chakra), happy to work inside an existing design system rather than rebuild from scratch
• Vitest / Jest, and the habit of testing the parts that matter
• Comfort in a large, lived-in codebase. You can read, navigate, and ship in 400k+ lines of someone else's TypeScript

Nice to have

• Mastra (agent/workflow framework), Vercel AI SDK v6
• Self-hosting Docker services (Compose, Caddy, basic Linux ops)
• TanStack Table, Tiptap, Zustand, Recharts
• Recruitment / sales / CRM domain knowledge
• Langfuse or other LLM observability tooling

How we work, and who fits

• Solo ownership. Async, no team to fall back on. You're comfortable being the engineer.
• Push-to-main discipline. We ship straight to main, several times a day. That takes rigour: small commits, real testing, and knowing when to pause.
• Pragmatic over pure. Bugs get fixed where they are, refactors are scoped, and we don't build for hypothetical futures.
• High AI fluency. You use Claude Code / Cursor / similar to move fast, and you read what they produce before it ships.
• Direct communication. You'll work directly with a non-engineering founder and can translate trade-offs into plain English.

What this is not

• Not green-field. You're inheriting and extending, not rebuilding.
• Not a team role. No other engineers to pair with or hand off to.
• Not a 9-to-5 ticket shop. Small business, real ownership, real impact, and real on-call when things break.

How to apply

Send a short note on why this role fits you, plus anything that shows how you work: GitHub, a system you've owned solo, an LLM or background-job project you're proud of. We care more about what you've shipped and owned than about a polished CV.