Description
Summary We're looking to validate a concept for an agentic AI automation system before committing to a full build. This is a POC engagement. The goal is a working prototype that demonstrates feasibility across a few high-impact AI use cases, not a production-ready system. The POC will cover a mix of practical AI automation scenarios, such as: Document/data extraction: LLM-powered extraction and classification from unstructured documents (invoices, forms, emails) Email inbox automation: an AI agent that reads, classifies, drafts, or routes emails automatically (e.g. auto-categorizing inquiries, drafting replies, flagging priority items) WhatsApp auto-responder: a conversational AI agent that can handle incoming WhatsApp messages, answer common queries, and escalate to a human when needed Agent-based workflow orchestration: tying the above together with agentic decision-making (routing, tool-calling, human-in-the-loop where relevant) You don't need to build all of these in full. We want to see a focused demonstration of feasibility on the use case(s) that best show off the approach, plus a recommendation on how the others could be layered in. We need someone who can move fast, make pragmatic technical decisions, and deliver a functioning demo. No over-engineering. What you'll work on: Building a lightweight LangGraph/LangChain agent workflow for the chosen use case(s) (ReAct or simple tool-calling pattern) LLM-powered extraction and classification with structured outputs (Pydantic/JSON schema) Integrating messaging channels (e.g. Gmail/Outlook API for email automation, WhatsApp Business API/Twilio for the auto-responder) A minimal FastAPI backend to expose/demo the workflow Integrating one LLM provider (OpenAI, Azure AI Foundry, or Anthropic) Basic prompt engineering and output validation (production-grade guardrails not required at this stage) Delivering a short demo/walkthrough (recorded or live) plus a brief write-up of feasibility, limitations, and recommended next steps Must-have skills: Strong Python Hands-on experience with LangGraph or LangChain (or equivalent agent frameworks) Experience with LLM APIs and prompt engineering FastAPI or similar lightweight backend framework Experience integrating messaging/email APIs (Gmail/Outlook Graph API, WhatsApp Business API/Twilio, or similar) Comfortable working in a fast, exploratory POC setting (not a long production sprint) Nice-to-have: Azure experience (AI Foundry, VMs, Power Automate) Document processing / OCR / extraction experience Experience building chatbots or conversational agents (WhatsApp, Teams, Slack) Prior POC or MVP delivery experience for AI products Deliverables: Working prototype (code + brief setup instructions) Short demo of the agent workflow(s) in action 1-page summary: what worked, what didn't, what a production version would need