AI Engineer Intern

AI Engineer Intern (Agent Architect) VectorStream
About Xepmi AI
Xepmi AI is building Explainable Persistent Machine Intelligence: AI systems that remember context over time and can show the reasoning behind every decision they make. Our flagship product, VectorStream, is an agentic go-to-market automation suite that turns scattered lead data, buyer signals, and sales conversations into a pipeline that runs itself. It captures leads, detects intent, engages prospects across WhatsApp, email, and AI voice, and traces every outcome back to the signal that caused it. We're starting with real estate in India as our beachhead and building for global markets from day one.
We're hiring an AI Engineer intern who wants to build production agentic systems, not just demos. You'll work directly with the founder on the core of VectorStream and ship features that real sales teams use every day.
What You'll Work On
Design and build agentic workflows across VectorStream's four-stage pipeline: Capture → Signal → Engage → Trace.
Build and orchestrate multi-agent systems with LangGraph, including planning, tool use, memory, state management, and human-in-the-loop checkpoints.
Build and maintain automation workflows in n8n that connect lead sources, CRMs (HubSpot, Zoho), WhatsApp, email, and internal APIs.
Integrate LLMs (Anthropic Claude, OpenAI, Gemini, and open-source models) into production services via APIs and SDKs.
Build tool integrations using the Model Context Protocol (MCP) so agents can securely act on external systems.
Develop RAG pipelines and vector search (FAISS, Chroma, Qdrant, pgvector) over property data, conversation history, and sales playbooks.
Work on AI voice agents using Vapi, with telephony via Twilio/Exotel, including call flows, latency tuning, and post-call summaries.
Implement guardrails, evals, and observability (e.g. Langfuse, LangSmith) so agents stay reliable, explainable, and cost-efficient.
Build compliance into the pipeline, including TRAI DND scrubbing, consent tracking, and audit trails for every automated action.
Use AI coding tools (Claude Code, Cursor) as part of your daily workflow to ship faster without cutting corners on quality.
Write clean, documented code and help set up deployment and CI/CD for AI services.
Build AI systems for:
Lead capture and enrichment
Buyer intent and signal detection
Lead scoring and prioritization
Multi-channel conversational follow-ups
Sales attribution and pipeline analytics
What You Bring
Strong Python skills and a working understanding of how LLMs, embeddings, and retrieval work
Hands-on projects with LLMs beyond simple API calls, such as a RAG app, an agent, or an automated workflow you actually built and can show us
Familiarity with at least one agent framework (LangGraph, LangChain, CrewAI, AutoGen, or similar)
Comfort working with REST APIs, webhooks, and JSON, since most of our system is things talking to other things
Curiosity about workflow automation tools like n8n, Make, or Zapier
The instinct to ask "what happens when this fails?" and to think about latency, cost, and reliability, not just whether it works once
Tech You'll Touch
LangGraph / LangChain · n8n · MCP · Claude / OpenAI / Gemini APIs · Vapi · Twilio / Exotel · WhatsApp Business API · FastAPI · FAISS / Chroma / Qdrant · PostgreSQL / Supabase · HubSpot / Zoho CRM · Langfuse / LangSmith · Docker · Git
Nice to Have
Experience building voice agents or conversational AI
Built something with n8n or another automation platform
Exposure to CRM systems, sales funnels, or marketing analytics
Experience with evals, tracing, or LLM observability tools
Open-source contributions, a GitHub full of side projects, or a hackathon win
Interest in real estate tech or B2B SaaS
Why Xepmi AI
Real ownership. You won't be one of fifty engineers. Your code goes into production and you'll own entire workflows end to end.
Work directly with the founder. Fast decisions, no layers of approval, and full visibility into how a product and company are built from scratch.
Modern stack. You'll work with the tools defining agentic AI right now, not legacy systems.
Hard, real problems. Making autonomous agents reliable, compliant, and explainable in live sales conversations is still unsolved territory.
Built India-first, for the world. You'll help shape a product designed to scale globally.
Growth path. Strong performers will be considered for a full-time role as the team grows.
Details
Duration: [e.g. 3–6 months]
How to Apply
Send your resume, GitHub, and a short note about the most interesting AI or automation project you've built to info@xepmi.ai with the subject line "Agent Architect Intern – [Your Name]."

Work arrangement
No