AI Intern

Vestval Group; India; On-site; INR 1,500,000 - 3,000,000; Full time

Job summary

Vestval Group is seeking an engineer to build and operate LLM-powered workflows, agents, and retrieval systems that weave into client and Vestval products. You will evaluate performance, prototype features with engineers, and document failure modes with mitigations.

You’ll work on end-to-end AI features across Vestval AI, Learn, and Flow, contributing to scalable, production-ready solutions.

Qualifications

• Working understanding of LLMs, embeddings and prompting.
• Comfort reading API docs and model cards.

Responsibilities

• Hands-on work with LLM systems in production: prompt and agent design, retrieval pipelines, evaluation, and integration of AI into Vestval products and client workflows.
• Build and test LLM-powered workflows and agents.
• Construct retrieval pipelines (chunking, embeddings, ranking).
• Run structured evaluations — measure quality, not vibes.
• Prototype AI features alongside product engineers.
• Document failure modes and propose mitigations.
• Help integrate AI features into Vestval AI, Learn or Flow.

Skills

LLMs

Embeddings

Prompting

API docs

Tools

pgvector

Pinecone

Qdrant

Job description

Hands‑on work with LLM systems in production: prompt and agent design, retrieval pipelines, evaluation, and integration of AI into Vestval products and client workflows.

• Build and test LLM-powered workflows and agents
• Construct retrieval pipelines (chunking, embeddings, ranking)
• Run structured evaluations — measure quality, not vibes
• Prototype AI features alongside product engineers
• Document failure modes and propose mitigations
• Help integrate AI features into Vestval AI, Learn or Flow

Required skills

• Working understanding of LLMs, embeddings and prompting
• Comfort reading API docs and model cards

Preferred / bonus

• Built a non‑trivial RAG system or agent end‑to‑end
• Experience with eval harnesses (Ragas, promptfoo, custom)
• Comfort with at least one vector store (pgvector, Pinecone, Qdrant)
• Statistical literacy — confidence intervals, sample sizes, bias awareness

Who should apply

People who have built something with an LLM API beyond a chat wrapper — an agent, a RAG system, an evaluation harness — and can explain its failure modes.

• You distrust demos that aren't reproducible
• You think in eval suites, not anecdotes
• You can explain why your system fails as clearly as why it works
• you treat hallucinations as a design problem, not a model defect

What you'll learn

• Agent design patterns and their real failure modes
• Evaluation methodology beyond accuracy numbers
• Safety, privacy and cost trade-offs in deployed AI

Tools & stack

Growth path

Pipeline into our applied AI engineering team, with exposure to enterprise AI deployments across BFSI, manufacturing and healthcare clients.

Months 0–3 (Intern)

Own evaluation for one production AI workflow. Ship one user‑facing AI feature.

Months 3–6 (Conversion)

Lead a small AI workflow end‑to‑end with evals, prompts and observability.

Own an AI surface in Vestval AI, Flow or a client engagement.

Years 2–3

Senior applied AI engineer — choose track: agents, retrieval, evaluation or platform.