Website:
geniffy.com
Job details:
We're Hiring: Backend Engineer — AI Integrations (Connected Apps / MCP Layer)
Geniffy | Remote | 2–8 Years Experience
Geniffy is an AI assistant that records your meetings, extracts action items, and actually gets them done — tickets created, emails drafted, recaps sent, calendar booked. We're not a notetaker. We do the work.
We're looking for one backend engineer to own the most important layer of our product: the Connected Apps / MCP integration system. This is where users connect their Gmail, Slack, Jira, Linear, Notion, and more then chat with Geniffy to take actions across all of them.
This layer is our biggest reliability gap and our biggest opportunity.
What you'll own
You'll be responsible for the reliability and intelligence of our cross-app execution layer — the system that turns a user's natural language request into real actions across their connected tools.
That means building and maintaining the integration infrastructure (Composio Sessions, OAuth flows, tool-call orchestration), ensuring every action Geniffy takes is accurate and verifiable (resolve-then-act patterns, confirmation workflows, idempotent operations, clean error handling), designing the entity resolution layer so the system correctly links people and objects across tools, and instrumenting the full pipeline with failure telemetry and a structured eval harness targeting ≥85% accuracy across 200+ scenarios.
When Geniffy says "ticket created" or "email drafted" — your work is the reason it's actually true.
Must have
→ 2–8 years backend engineering (Python / FastAPI strongly preferred) → Production experience with LLM tool-calling or function-calling (OpenAI, Anthropic, or similar) → Worked with at least one of: MCP, Composio, LangChain agents, or Zapier-style integration platforms → Strong on OAuth 2.x, Redis, Postgres, async patterns → Comfortable in React / TypeScript for chat-side UI — connection flows, confirmation cards
Nice to have
→ Prompt engineering depth — system prompts, agent reliability, structured retries → Experience building eval harnesses or AI observability → Has shipped a multi-app AI chat product
The profile
Hands-on IC who ships code daily. Comfortable working with ambiguity when the problem is open, and executes cleanly when handed a clear spec. You care about whether the thing actually works in production, not just in a demo.
Why this role matters
Every AI meeting tool stops at notes. Geniffy doesn't. After the meeting ends, it creates the tickets, drafts the emails, sends the recap, schedules the next sync across whatever tools the team uses.
The Connected Apps layer is how all of that happens. You'd be the engineer who makes it reliable.
Click on Apply to know more.