Overview

Onsite: Gurgaon | 6 Days Working

Responsibilities

• Design and build RAG pipelines — chunking, embeddings, vector search, re-ranking
• Build agents and orchestration logic from scratch (no framework crutch) — tool-use, multi-step reasoning, state management
• Train and fine-tune models where off-the-shelf LLMs fall short — dataset curation, fine-tuning, LoRA, evaluation of trained models
• Set up eval harnesses and benchmarks to catch regressions before users do
• Implement guardrails, hallucination detection, and prompt-injection defenses
• Optimize for cost, latency, and context-window efficiency — caching, streaming, Redis
• Design backend APIs and data models independent of the AI layer
• Containerize and deploy services with Docker/Kubernetes; build MLOps pipelines for model versioning, monitoring, and rollout
• Run A/B tests and iterate on prompt/model performance
• Maintain observability and tracing across LLM pipelines

Qualifications

• Important: We’re specifically looking for someone who has spent their experience building and shipping real AI agents, agentic systems, or RAG systems used by real users at scale.This is not a general software engineering role. If your experience is primarily in backend/software engineering and you’ve only recently started working with LLMs or AI, please don’t apply.
• We’re looking for someone whose ~2+ years of experience is predominantly hands-on AI engineering - building production agents, RAG pipelines, LLM systems, orchestration, evaluation, and related infrastructure.
• Tech Stack: Python, LLM APIs, RAG, Vector DBs, Model Training/Fine-tuning, Docker, Kubernetes, Redis, MLOps, FastAPI, Postgres, AWS
• About the RoleBuild and ship production LLM systems - not prototypes. You'll own the full lifecycle: retrieval architecture, agent/orchestration design, model training/fine-tuning, evaluation, and deployment for features with real users and real failure consequences.
• 2+ years of hands-on experience building and shipping production AI/LLM systems
• Strong experience building AI agents, agentic workflows, or RAG systems used by real users at scale
• Strong backend fundamentals — API design, DB modeling, Python
• End-to-end RAG fluency — embeddings, vector DBs, re-ranking
• Ability to design and build agent/orchestration systems from first principles, without relying on frameworks like LangChain/LangGraph
• Hands-on experience with model training/fine-tuning — LoRA, dataset curation, evaluation
• Docker, Kubernetes, and MLOps practices — CI/CD for models, versioning, monitoring
• Redis or similar for caching/session state