AI Engineering Intern (Python)

AI Engineering Intern (Python) at Valura.AI

Position: AI Engineering Intern (Python)

Company: Valura.AI

Locations: Chennai, GIFT City (Gandhinagar), or Remote

Duration: 3 to 6 months (Full-time preferred)

Compensation: Competitive stipend based on experience

Career Path: High performers may transition to full-time AI Engineer roles

Company Overview

Valura.AI is a regulated global investing platform operating from GIFT City and the UAE. We build the cross-border investment infrastructure that lets people in India and the UAE invest globally, under real regulatory and accuracy constraints.

Our AI layer is not a chatbot bolted onto a website. It sits on top of live portfolio data, market data and a double-entry ledger, and it answers questions where being confidently wrong is a real problem.

About the Role

You will work on our AI microservice: a Python multi-agent system that powers in-app research, portfolio explanation and customer support across our products.

This is production work, not a toy project. Whatever you build gets used by real users looking at their real money. That is the fun part and also the hard part: in finance, a plausible-sounding wrong answer is worse than no answer. A lot of the job is building the guardrails, evaluations and retrieval that keep the system honest.

You will not be handed a research paper and told to reproduce it. You will be handed a real problem, a real dataset, and a fair amount of ownership.

Tech Stack

Python 3.11+

FastAPI

LLM APIs (Claude, GPT) and structured tool calling

Multi-agent orchestration and MCP tool servers

RAG: embeddings, vector search, retrieval pipelines

PostgreSQL and SQLAlchemy

Redis

Docker

AWS

Git

What You'll Work On

Build and extend agents in our multi-agent orchestrator: research, portfolio analysis, market commentary, support.

Design tools and MCP integrations that let agents pull real data (portfolios, market feeds, ledger balances) instead of guessing.

Write and improve prompts as versioned, tested artifacts, not one-off strings.

Build retrieval pipelines over fund documents, factsheets, research notes and regulatory content.

Build evaluation harnesses: golden datasets, regression suites, LLM-as-judge scoring, so we know when a change makes things worse.

Add tracing and observability so we can see exactly why an agent said what it said.

Work on guardrails: grounding checks, refusal behaviour, and making sure the system never invents a number.

Integrate the AI layer with our chat surfaces (in-app, WhatsApp, Telegram) and the APIs behind them.

Optimise for latency and cost: caching, model routing, prompt size, streaming.

Ship to staging and production with Docker on AWS, and help debug it when it misbehaves.

What We're Looking For

Required

Solid Python. You can read someone else's code and extend it without rewriting it.

Hands-on experience calling LLM APIs, even if only from side projects. You have actually built something, not just read about it.

Basic SQL and comfort with data: joining, cleaning, sanity-checking.

Comfort with Git and the terminal.

Clear thinking and clear writing. Much of this job is precisely describing what you want a model to do.

Genuine skepticism. If a model gives you an answer, your instinct should be to verify it.

Nice to Have

RAG or vector-database experience (pgvector, FAISS, Qdrant, anything).

Agent frameworks, tool calling, or MCP.

Evaluation experience: you have measured model quality rather than eyeballing it.

FastAPI, Docker, or any cloud deployment exposure.

Interest in markets and investing. You do not need finance experience, but curiosity helps a lot.