AI Engineer

AI ENGINEER

Location: Mumbai

Experience: 3-5 years, with at least 2 years building LLM systems that

ran in production

ABOUT THE ROLE

You will own an AI service from the retrieval code through to the token bill it generates. We build agentic and RAG systems for clients in e-commerce, healthcare, financial services and enterprise SaaS. These run under real load with real money attached, so the job splits roughly intobuilding the system, proving it behaves correctly, and keeping it fast and affordable.This is a senior individual contributor role. You will review other people's code and carry associate engineers.

WHAT YOU WILL DO/OWN

• Build multi-agent and RAG systems on LangGraph, with agent memory,tool permissioning, spend caps and human approval on anythingirreversible.

• Design the retrieval layer and be able to say what its recall and precision are, separately from answer quality.

• Own evaluation. Build golden sets from production traffic, run them in CI, and decide what score change blocks a release.

• Instrument tracing and monitoring so that when a client says the answer was wrong, you can find the request and explain why.

• Run the serving layer. Latency budgets, caching, batching, model routing, and a cost per request you can quote without looking it up.

• Defend the system against prompt injection, data leakage across tenants, and jailbreaks, and test those defences rather than assume them.

• Integrate with client systems: Salesforce, SAP, ServiceNow, Magento, Odoo, and whatever else the account runs on.

• Work directly with client engineering teams during delivery.

WHAT YOU NEED

Engineering: Advanced Python with async and typing, Production API design

in FastAPI or Django REST, SQL and data modelling, Docker, Git and code review, Linux debugging, testing systems whose output changes between runs.

Models: OpenAI, Anthropic and Gemini APIs hands on. Amazon Bedrock, Azure

AI Foundry or Google Vertex AI. STT/TTS. Vision and image models including CLIP, SAM, Stable Diffusion or FLUX. Tool calling with schema enforcement, structured output, streaming, and context and token budget management. Context Engineering

Agents & orchestration: LangGraph and LangChain, plus one of CrewAI, AutoGen, OpenAI Agents SDK or Pydantic AI. MCP. Idempotency and compensating actions for tools that change state.

Retrieval & knowledge: RAG, chunking strategy, embedding selection, hybrid search, reranking, permission-aware retrieval, document parsing. PostgreSQL with pgvector and one dedicated vector database at scale (Qdrant, Weaviate, Milvus or Pinecone), GraphRAG with Neo4j.

Evaluation and safety: Ragas or DeepEval, LLM as judge with calibration, regression gates in CI, LangSmith or Langfuse, OpenTelemetry, drift monitoring, PII redaction with Presidio, prompt injection defence.

Serving & performance: Latency budgeting — p50 / p95 / p99, Semantic caching, batching, model cascade and routing, and cost instrumentation.

Machine learning: scikit-learn, XGBoost, Feature engineering, Time series — ARIMA, Prophet, LSTM, Hugging Face Transformers, PEFT with LoRA or QLoRA, and judgement to know when a classical model beats an LLM.

Good to have: WebRTC and LiveKit for voice, vLLM / TGI / Triton, NeMo Guardrails / Llama Guard, Open-weight models such as Llama, Mistral or Qwen, PyTorch, OpenCV, Semantic Kernel

Pay: ₹900,000.00 - ₹1,200,000.00 per year

Benefits:

• Health insurance
• Leave encashment
• Life insurance
• Provident Fund
• Work from home

Work Location: In person