AI Engineer

REVERONIX; Bengaluru; On-site; INR 1,200,000 - 2,400,000; Full time

Job summary

REVERONIX is seeking a software engineer to design and ship AI-powered features in client products from Bengaluru. You will build evaluation harnesses and treat agent quality as a first-class metric, while working across LLM providers like Anthropic and OpenAI and various embedding stacks.

You will pair with backend engineers on integration, observability, and fallback design. The role requires 3+ years in software engineering, strong Python (or TypeScript), and hands-on experience with vector

Qualifications

• 3+ years software engineering with production AI experience.
• Proficiency in Python or TypeScript.
• Experience with vector stores, embeddings, and retrieval-augmented generation (RAG).
• Experience building and running evaluations and guardrails for models.

Responsibilities

• Design and ship AI-powered features into client products.
• Build evaluation harnesses and treat agent quality as a first-class metric.
• Work across LLM providers (Anthropic, OpenAI) and embedding stacks.
• Collaborate with backend engineers on integration, observability, and fallback design.

Skills

Python

TypeScript

AI systems

Vector stores

Embeddings

RAG patterns

Evaluation

Job description

• Design and ship AI-powered features into client products
• Build evaluation harnesses and treat agent quality as a first-class metric
• Work across LLM providers (Anthropic, OpenAI) and embedding stacks
• Pair with backend engineers on integration, observability, and fallback design

What you'll do

• Design and ship AI-powered features into client products
• Build evaluation harnesses and treat agent quality as a first-class metric
• Work across LLM providers (Anthropic, OpenAI) and embedding stacks
• Pair with backend engineers on integration, observability, and fallback design

What we look for

• 3+ years software engineering, with at least 1 year shipping AI systems in production
• Solid Python (or TypeScript) and comfort with vector stores, embeddings, RAG patterns
• Have built and run evals — not just prompts in a notebook
• Honest about model limits; designs with guardrails