Staff AI Engineer

ChargePoint operates one of the largest EV charging networks, and this role builds the AI systems layered on top of its monitoring and analytics platform. The brief spans voice AI, computer vision and conversational AI, applied to customer support and to automation across the charging infrastructure itself. In practice this is production LLM engineering: LangChain and LangGraph, Amazon Bedrock and OpenAI APIs, RAG architectures, vector databases, and Elasticsearch used for hybrid search combining BM25 with dense embeddings. The posting emphasises low latency inference, caching strategy and observability, which suggests they have already been burned by prototypes that did not survive production. It reports to the Director of NOC Delivery and Automation, so the work is tied to network operations rather than research.

What the posting requires: 8 or more years of software engineering experience with 4 or more focused on AI and ML systems in production. A B.Tech, B.E or M.S. in computer science, machine learning or a related field, which ChargePoint specifies should be from what it calls a top tier institution. Deep expertise in Python, FastAPI, Django and modern backend frameworks. Hands on LLM engineering: LangChain, LangGraph, Bedrock or OpenAI APIs, prompt engineering and RAG. Strong Elasticsearch including vector search, hybrid search and semantic retrieval. Vector databases such as Qdrant, ChromaDB or Pinecone. Production LLM systems with low latency inference, caching and observability. Distributed systems design, microservices and event driven patterns with Kafka or RabbitMQ. Cloud (AWS or GCP), Docker, Kubernetes and CI/CD. Strong PostgreSQL including query optimisation and schema design, plus Redis and MongoDB. A track record of measurable performance improvements.

Also listed: voice AI and telephony integration (Genesys, SIP) and speech pipelines; computer vision and visual transformers; data engineering with Airflow, DBT, ClickHouse, Trino and Iceberg; experience with datasets over 50TB; fine tuning with LoRA or QLoRA; RLHF style preference tuning; observability stacks; and certifications in generative or agentic AI.

Location and office reality: the requisition header says India and the posting body states Bangalore, India.

Who this is for: an engineer who has already put LLM systems into production and can talk about latency and cost rather than model choice. The requirement list is long and reads like several roles fused together, so expect to be strong in the LLM and search half and passable on the voice and vision half rather than expert in all of it.