AI Engineer — Prompt‑Driven Development
Claude Code wrote code in 10 seconds. It took 2 hours to fix what it broke. If that sounds familiar, you’ll get why we exist.
We believe the current approach is backward. Patching code via “agentic” chats is like throwing a grenade into a codebase — you don’t control what’s added, deleted, or duplicated, and you’re left cleaning up. Prompt‑Driven Development (PDD) flips the model: the human‑readable prompt becomes the durable source of truth, and code is the generated artifact. Update intent in one place, regenerate clean code, tests, and docs in batch, and keep everything in sync.
We’re building the open‑source toolchain (pdd‑cli) and platform to make this real, starting with high‑impact migrations (e.g., Heroku/Rails → AWS/Next.js) delivered in weeks with 70%+ cost savings.
Why PDD (Now)
– Pain is real: Fast AI writes; humans spend hours fixing drift and regressions.
– Prompts as source: Intent lives in `.prompt` files — versioned, reviewable, auditable.
– Batch over babysitting: Deterministic, reproducible runs replace ad‑hoc chat poking.
– Full slice generation: From one prompt → code, tests, examples, docs, IaC stubs.
– Compounding context moat: Each migration yields few‑shot examples that improve the next.
See the whitepaper for the deeper case and workflow: https://github.com/promptdriven/pdd/tree/main/docs/whitepaper_with_benchmarks.
Role Overview
– Title: AI Engineer
– Engagement: Part‑time trial/full‑time.
– Location: On‑site in Palo Alto, CA (in‑office collaboration).
– Impact: Build the primitives that make regeneration trustworthy at scale — then ship real migrations on top.
What You’ll Champion
– PDD workflow primitives: Implement `generate`, `example`, `test`, `fix`, `update` with back‑prop so prompts/code/tests stay synchronized.
– Batch LLM pipelines: Orchestrate retries, sharding, caching, cost controls; target multi‑provider batch APIs.
– Verification layers: Crash‑to‑pass loops, regression suites, spec checks; accumulate tests as the safety net.
– Developer experience: CLI flows, editor/MCP integrations, example marketplace submission and retrieval.
– Productionization: Infra as code (Terraform/CDK), blue‑green + rollback, observability, CI hooks.
What You Bring
– Early Stage Startup mindset
– Strong software fundamentals; shipped production systems end‑to‑end.
– Hands‑on LLM experience: prompt design, few‑shot assembly, retrieval/context tooling, evals, and cost/perf trade‑offs.
– Reliability mindset: treat determinism as a feature; design graceful fallbacks when it isn’t possible.
– Comfort across boundaries: backend services, CI pipelines, and developer UX for CLI/editor.
– Bonus: cloud migrations, Terraform/CDK, Next.js, Firebase Functions, formal/spec‑based testing, ranking/marketplaces.
Tech Stack
– Backend: Python 3.12 (Firebase Functions), pytest, batch orchestration.
– Frontend: Next.js (App Router, TypeScript), component examples.
– Cloud/Infra: AWS/GCP/Vercel, blue‑green/rollback, observability, CI.
– LLMs: PDD, Multi‑provider batch APIs; few‑shot retrieval and scoring/eval harnesses.
Compensation
– Early‑stage package with a large **equity** component; open source contributor volunteer to start, with a clear path to full‑time and refreshed equity on conversion. We believe in building strong team bonds so we provide snacks, lunch and dinner.
Why Join
– Own the paradigm: Help define prompt‑first engineering where code is generated, not patched.
– Compounding advantage: Every migration adds few‑shot examples that boost the next run.
– High leverage: Small, senior team; you’ll own core architecture and ship visible wins.
– Right timing: Model quality, context windows, and batch economics just unlocked full‑system generation.
Make your application standout by:
- Staring⭐️ our GitHub Repo:https://github.com/promptdriven/pdd
- Joining our Discord: https://discord.gg/Yp4RTh8bG7
Application Process:
- Phone Screen
- Interview
- Trial period as a Open Source Contributor (volunteer) to PDD
- Full-time employment