Full Stack Engineer

Full Stack Engineer

San Francisco based start-up
Delamain AI is an intelligence platform serving middle-market firms. This engineer owns the full stack — from real-time data pipelines and LLM summarization engines to the user-facing web portal.

Required Skills

Languages

TypeScript (Expert) — all production frontend/backend code
Python (Proficient) — pipelines, scraping, LLM orchestration
SQL / MongoDB (Proficient) — BigQuery ingestion, aggregation pipelines
Bash/Shell — EC2 ops, cron, deployment scripts

Frameworks & Infrastructure

Next.js (App Router), React, Tailwind CSS / shadcn/ui, NextAuth.js
Node.js + Server Actions , OpenAI-compatible API (streaming, tool calling)
MongoDB Atlas — document modeling, Atlas Search
Docker / Docker Compose — all services containerized
Linux / tmux — SSH, process management, EC2 service administration

Engineering Capabilities

Full stack ownership: data model → API → UI → deploy (no tier handoff)
Async/concurrent systems & data pipeline architecture (ETL/ELT, deduplication)
Web scraping at scale — anti-bot evasion, government/political sources Playwright,BeautifulSoup)
Real-time streaming — SSE/WebSockets for agent responses
Production debugging across multi-stage pipelines
Multi-tenant security — data isolation, access control, audit logging

AI / LLM

LLM API integration — chat completions, tool/function calling, streaming, context management
Prompt engineering for summarization, personas, research agents
Agent architecture — multi-step tool use, memory patterns, RAG
AI output evaluation — hallucination detection, guardrails

Preferred Skills

Rust — Core system runs on Rust
AWS Bedrock — Claude 3.x families; cost/quality/latency tradeoffs AWS (EC2, S3), Cloudflare (DNS, tunnels), CI/CD (GitHub Actions)
Google BigQuery — large-scale parallel data ingestionRedis — caching, job queuing, pub/sub
Signal CLI — secure messaging integration on Linux EC2
Multi-provider LLM routing (Fireworks AI or similar)
Embedding + vector search (Voyage AI or equivalent)
Automated briefing pipelines — scheduled ingest → summarize → deliver