AMPITY INFOTECH

Job Description

AI Engineer

LLM / GenAI Systems | ML & Model Engineering | Applied AI Delivery

Location: Remote (India)

Employment Type: Full time

Team: Client Delivery Engineering

Reports To: Tech Lead / Engagement Lead

About Ampity Infotech

Ampity Infotech is a cloud and DevOps engineering firm that helps growing companies design, build, and operate production-grade software and infrastructure on AWS. We work directly with client engineering and leadership teams, and we build AI-powered capabilities into client products as a core part of our delivery, not as a side experiment.

Role Summary

As an AI Engineer on Ampity's delivery team, you will design, build, and ship AI-powered capabilities for client products. The work spans applied LLM/GenAI engineering (RAG pipelines, agentic workflows, prompt design, evaluation) and classical ML/data science (model training, fine-tuning, feature engineering, deployment) depending on what an engagement calls for. You will be accountable for how these systems perform in production, meaning accuracy, latency, cost,



and reliability, not just for a working prototype.

What You'll Own
- Design and build LLM/GenAI-powered features: RAG pipelines, agentic or tool-calling workflows, prompt design, and integration with model providers (OpenAI, Anthropic, AWS Bedrock, etc.).
- Build, fine-tune, and evaluate ML models where a classical or statistical approach fits the problem better than an LLM, including data preprocessing and feature engineering.
- Design and run evaluation frameworks (offline evals, golden datasets, human-in-the-loop review) to measure accuracy, hallucination rate, and regressions before and after changes.
- Deploy and operate AI/ML systems in production on AWS (SageMaker, Bedrock, Lambda, or EKS), with monitoring for drift, latency, and cost.
- Own data pipelines that feed models and retrieval systems: ingestion, chunking, embeddings, and vector store management.
- Apply security and responsi

📌 AI Engineer (Pune)
🏢 Ampity Infotech
📍 Pune