About the role
Responsibilities:
- Design and ship AI-powered product features (LLMs, RAG, agents, ML APIs) into our existing services, working closely with backend, frontend, and data science teams.
- Integrate off-the-shelf and in-house models (LLMs, embeddings, ML APIs) into robust microservices and user-facing flows.
- Design and implement RAG and workflow/agent pipelines: retrieval, context assembly, tools integration, guardrails, and fallbacks.
- Own AI service reliability in production: latency, throughput, cost, observability, circuitbreakers, and rollback/versioning of models and prompts.
- Collaborate with Data Scientists who own model training/finetuning and evaluation design; productionize their outputs as stable APIs/workflows.
- Implement logging, feedback capture, and lightweight online evaluation hooks to measure the quality of AI features over time.
- Ensure safety, security, and compliance for AI features: prompt injection defenses, PII handling, abuse/hallucination controls, and audit trails.
- Contribute to internal AI tooling: SDKs, templates, and reusable components to accelerate future AI use cases.
Requirements:
- Skills required: AI Engineer role demands more than AI-based augmentation with in-depth understanding of concepts like- -RAG -GenAI, LLM finetuning, Prompt engineering, Multi-Agent framework (langchain, langraph etc with hands on experience), Eval generation and their importance, Tokens usage and optimisations, Model/mcp gateway.
- Strong software engineering in Python (and one of Node/Java/Go), REST/gRPC APIs, queues, and microservices on cloud infrastructure.
- Hands-on experience shipping at least one AI powered product to production (e. g., search, recommendations, chatbots, summarization, classification).
- Practical knowledge of LLM concepts: prompts, context engineering, embeddings, vector search, basic evaluation metrics, and latency/cost trade-offs.
- Solid understanding of integration patterns with third-party AI providers (OpenAI, Anthropic, etc. ) and vector DB.
- Hands-on and good understanding of at least one agentic framework like Langgraph.