Website:
devaisuite.com
Job details:
AI Solutions / Integrations Engineer (Forward-Deployed)
DevAI Suite is an AI-native B2B SaaS platform for manufacturing programme management (APQP/PPAP/FMEA; automotive, aerospace, general manufacturing). We build custom packages for individual manufacturers — bespoke integrations, customer/ordering portals, and custom AI agents that connect their systems and workflows to our platform.
We're looking for an engineer who is three things at once: a strong systems integrator, a hands-on AI/agent builder, and a customer-facing problem-solver. You'll own custom work end-to-end and be the technical person our customers trust.
Stack: Python 3.11 / FastAPI · Next.js / React / TypeScript · PostgreSQL (multi-tenant, row-level security) · Fly.io · REST APIs, OAuth2, webhooks · LLM tool-use / MCP · RAG (pgvector, Voyage/Together embeddings) · fine-tuned models.
What you'll do
Build custom connectors integrating customer systems (SAP, MES, ordering/ERP, EDI, CSV/SFTP) with the DevAI Suite API — auth, data mapping, sync, error handling.
Stand up per-organization customer/ordering portals on top of our platform.
Build custom AI agents & automations for customers on our agentic fabric — tool-using agents that operate under RBAC + entitlement + propose→approve (human-in-the-loop), extending our MCP connector-builder and coordinator tools.
Stand up tenant-grounded AI — RAG over a customer's own documents, custom guidance/support assistants, prompt + eval tuning; where it pays off, model fine-tuning / adapters and routing.
Ship fast with AI-assisted development (Cursor / Claude Code) — and review the AI's output critically.
Be the front-line technical contact: diagnose integration, data, and AI-behavior issues, resolve or escalate, and communicate clearly under pressure.
Turn recurring fixes into reusable patterns; feed product gaps back with reproducible detail.
You have
Real experience integrating systems over APIs: REST, OAuth2 / API-key auth, webhooks, pagination, retries, idempotency, rate limits, data mapping/ETL.
Production AI/LLM engineering — you've built tool-using / function-calling agents, done RAG (embeddings + a vector DB), and can prompt-engineer with an evals + guardrails mindset (grounded answers, no hallucinated actions, tenant-safe retrieval). You understand why an agent misbehaves, not just how to call an API.
AI-assisted development fluency (Cursor / Claude Code / Copilot) — you ship quickly with AI in the loop and critically review what it generates. You don't paste code you can't defend.
Python (FastAPI a plus) and working JavaScript/TypeScript (React/Next helpful).
Comfort with SQL / relational data and a real instinct for multi-tenant data isolation.
Strong debugging across the stack — logs, network traces, DB, and non-deterministic AI behavior.
Excellent written and verbal communication. You can explain a technical (or AI) problem to a plant quality manager and stay calm during a production incident. (Non-negotiable — it's a third of the job.)
Bonus
Model work: fine-tuning / LoRA, model routing & serving (RunPod/Together/hosted), embeddings backends (Voyage/Together, pgvector), evals/benchmarks.
Agent frameworks & MCP (Model Context Protocol), multi-step orchestration, retrieval pipelines.
Manufacturing / ERP / MES / PLM domain (SAP, Teamcenter, Oracle, MES, QMS, EDI) — a major plus.
Cloud/containers (Fly.io, Docker), CI/CD; security awareness (secrets, tenant isolation, least privilege, AI safety).
Prior Solutions / Forward-Deployed / Implementation Engineer or senior technical-support background.
How we work: small, fast, staging-first, high ownership, AI in every loop. You'll have a connector framework, a portal template, an agentic tool fabric + MCP builder, a support runbook, and a docs source-of-truth to build on — not a blank page.
Click on Apply to know more.