Senior AI Engineer

Keka Technologies Private Limited; NTR; On-site; INR 3,500,000 - 7,000,000; Full time

Job summary

Keka Technologies Private Limited is seeking a senior AI/ML engineer to own the end-to-end LLM platform. You will drive the LLM Process Synthesizer, build agentic workflows with LangChain/LangGraph, and architect robust RAG systems using embeddings and vector databases.

You will optimize retrieval, evaluation frameworks, prompt management, and ensure scalable deployments on AWS with Docker and PostgreSQL stacks.

Qualifications

• Strong proficiency in Python and building production-grade LLM applications.
• Hands-on experience with Claude, OpenAI APIs, structured outputs, tool calling, and RAG.
• Experience with LangChain, LangGraph, or similar agent frameworks.
• Strong knowledge of embeddings, vector databases and retrieval design.
• PostgreSQL, SQLAlchemy, AWS, and Docker experience.
• Strong understanding of AI evaluation, guardrails, debugging, and system design.

Responsibilities

• Own LLM Process Synthesizer, turning events into grounded, verifiable process specs.
• Build agentic workflows with LangChain/LangGraph; include tool-calling and human-confirm flows.
• Design and optimize RAG systems using embeddings, semantic search, and vector DBs.
• Own retrieval architecture: chunking, indexing, embedding selection, and vector stores.
• Define AI evaluation frameworks, grounding checks, and replay verification.
• Deploy scalable AI solutions using AWS, Docker, and PostgreSQL-based stacks.

Skills

Python

Claude

OpenAI APIs

LangChain

LangGraph

Embeddings

Vector databases

PostgreSQL

SQLAlchemy

AWS

Docker

AI evaluation

Guardrails

Debugging

System design

Education

B.Tech/B.E. (CSE, IT, ECE) or equivalent

Tools

pgvector

Qdrant

Pinecone

Weaviate

PostgreSQL

SQLAlchemy

AWS

Docker

Job description

Responsibilities

• Own the LLM Process Synthesizer, transforming captured action events into grounded, verifiable process specifications using Claude, structured outputs, and tool-use capabilities.
• Build agentic workflows using LangChain/LangGraph, including tool-calling agents and propose-diff → human-confirm execution flows.
• Design, implement, and optimize RAG systems using embeddings, semantic search, and vector databases.
• Own retrieval architecture, including document chunking, indexing strategies, embedding selection, and vector stores (pgvector/Qdrant).
• Define and maintain AI evaluation frameworks, grounding checks, coverage/replay verification, and hallucination controls.
• Implement prompt/version management, autonomy gating, and cost/latency optimization.
• Deploy and maintain scalable AI solutions using AWS, Docker, and PostgreSQL-based systems.

Skills Required

• Strong proficiency in Python and building production-grade LLM applications.
• Hands‑on experience with Anthropic Claude, OpenAI APIs, structured outputs, tool calling, and RAG.
• Experience with LangChain, LangGraph, or similar agent frameworks.
• Strong knowledge of embeddings, vector databases (pgvector, Qdrant, Pinecone, Weaviate, etc.), and retrieval design.
• PostgreSQL, SQLAlchemy, AWS, and Docker experience.
• Strong understanding of AI evaluation, guardrails, debugging, and system design.

Qualifications

• B.Tech/B.E. (CSE, IT, ECE) or equivalent.
• Minimum 5+ years of relevant experience in Software Engineering, AI/ML, or Generative AI development.

Preferred

• Experience in Process Mining or Workflow Modeling.
• Knowledge of self‑hosted/on‑premise model serving.
• Experience with embedding fine‑tuning and retrieval optimization.
• Familiarity with prompt evaluation tools, AI observability, and model monitoring.
• Understanding of PII handling, data privacy, and security best practices.