Xcubator

Role in one line

Own the user-facing applications and the web / API layer of the agentic platform — the chat shell, the per-agent UIs, and the glue that connects them to the agent runtime — shipping production-quality interfaces at high velocity.

Context

We are building a multi-agent AI platform for a regulated banking client: a shared “spine” (chat shell, API gateway, agent runtime, tool layer) with several agents running on top — KYC assistant, document generator, meeting analysis, search, and office / e-mail productivity. The full-stack engineer owns the experience layer end to end.

What you will work on

• Customise and extend LibreChat (React / TypeScript) — branding, auth integration, and new agent-specific UI surfaces.
• Build polished, usable internal tools: slot-filling document-generation flows, KYC review / approval screens with human-in-the-loop gates, and meeting-analysis Q&A and task views.
• Implement streaming LLM output in the UI (SSE / websockets), draft → preview → correct → download flows, and stateful multi-step interactions.
• Wire the frontend to backend services — FastAPI (Python) agent endpoints and the Laravel gateway — including SSO via Azure AD / Entra ID (OpenID Connect).
• Move fast with AI-assisted development while keeping the codebase reviewable, typed, and maintainable for a regulated client.

Must-have

• Strong React + TypeScript: component architecture, state management, responsive and accessible UI.
• Real product / UX sensibility — you build internal tools people actually want to use, not just functional ones.
• Vibecoding fluency: highly productive with AI coding assistants (Claude Code, Cursor, Copilot) — able to scaffold, iterate, and ship quickly without losing control of code quality.
• Solid backend competence: consuming and integrating REST / streaming APIs, auth flows (OIDC / SSO), and comfort reading and extending Python and / or PHP services.
• Engineering hygiene: Git, CI/CD, and containerised local development (Docker).
• Codebase agility: comfortable working inside a large existing open-source codebase and bending it to requirements.
•  

Nice-to-have

• Prior LibreChat or chat-UI customisation experience.
• SSO integration experience with Azure AD / Entra ID (OpenID Connect).
• Hands-on Python (FastAPI) and / or PHP (Laravel), not just integration.
• Experience building UIs for LLM / agent products (streaming, tool-call surfacing, citations).
• Prior work in regulated, enterprise, or security-sensitive environments.

Tech stack you will touch

React, TypeScript, Tailwind / CSS, Node tooling, LibreChat · FastAPI (Python) and Laravel (PHP) services · Azure Entra ID / OpenID Connect SSO · Docker, Git, CI/CD · streaming (SSE / websockets).

Ways of working

• Remote, distributed delivery team; English working language; scrum-light cadence.
• Banking-grade rigor: every action auditable, human-in-the-loop by design, compliance considerations built into the architecture — not bolted on.
• Vibecoding is for velocity, not for shipping unreviewed code: AI-assisted, human-owned, review-clean.

Important

As this is a Germany-based project, we are primarily seeking candidates based in Western Ukraine, with Vinnytsia and Lviv being our preferred locations. 

Role in one line

Build the agent runtime — the brains of the platform: stateful LangGraph workflows, RAG, tool integrations, model routing, human-in-the-loop gates, and audit-grade logging — in Python.

Context

We are building a multi-agent AI platform for a regulated banking client. The agent runtime is a Python / FastAPIservice exposing OpenAI-compatible endpoints, with agents implemented as LangGraph graphs and tools exposed over MCP. Sensitive workloads (KYC, PII) run on self-hosted open-weight models on EU infrastructure (Hetzner); less sensitive workloads route to frontier models via AWS Bedrock EU. The AI engineer owns the agent logic, retrieval, and the model / tool layer.

What you will work on

• Design and build stateful agent graphs in LangGraph — multi-stage workflows, tool calling, and interrupt-based human-in-the-loop gates (e.g. the six-stage KYC assistant).
• Build RAG pipelines: ingestion, chunking, embeddings, retrieval, and grounded Q&A with source citations (e.g. meeting analysis, search lens).
• Implement structured-output patterns where the LLM emits validated JSON only and deterministic engines do the rest — keeping the model out of the render path for document generation.
• Integrate tools over MCP (streamable-http) and wire agents to bank data sources behind the tool layer.
• Work across hosted and self-hosted models — a per-task model router selecting between Bedrock EU and on-prem open-weight models — with prompt design, evaluation, and cost / latency awareness.
• Build audit-grade logging and provenance for every AI operation, with PII handled by hashing rather than plaintext.

Must-have

• Strong Python: production-grade, typed, tested — not notebook-only.
• LangChain and LangGraph: hands-on building stateful graphs, tool / function calling, and human-in-the-loop interrupts.
• LLM application patterns: RAG, prompt design, structured output / JSON-schema validation, and evaluation of agent behaviour.
• Retrieval stack: vector stores (pgvector and / or Qdrant) and embeddings.
• Serving: FastAPI for agent endpoints; comfort with async Python.
• Model access: working with hosted APIs (AWS Bedrock) and self-hosted open-weight models (Llama, Mistral, Qwen).

Nice-to-have

• Banking / regulated experience; awareness of audit logging, PII handling, and EU data residency.
• MCP (Model Context Protocol) tool integration.
• ASR / speech-to-text pipelines (for meeting analysis).
• MLOps for self-hosted inference (vLLM, Ollama, or similar) on EU infrastructure (Hetzner).
• Workflow orchestration (e.g. Camunda) and deterministic document rendering (docxtpl, openpyxl).

Tech stack you will touch

Python, LangChain, LangGraph, FastAPI · pgvector / Qdrant, embeddings · AWS Bedrock EU and self-hosted open-weight LLMs on Hetzner · MCP servers (streamable-http) · Docker, Git, CI/CD.

Ways of working

• Remote, distributed delivery team; English working language; scrum-light cadence.
• Banking-grade rigor: every AI operation logged and auditable, human-in-the-loop by design, compliance built into the architecture — not bolted on.
• Agents prepare, retrieve, draft, and propose; the trigger and the final action always stay with a human.

Important
As this is a Germany-based project, we are primarily seeking candidates based in Western Ukraine, with Vinnytsia and Lviv being our preferred locations.