EV.analytica gives electric vehicle drivers a mobile platform to plan routes, find charging stations, monitor vehicle health, and manage charging costs. Our challenge was to prove that a production-grade AI agent could sit on top of their existing data and APIs, letting users ask questions and trigger actions in plain language, by text or voice.
EV.analytica's platform already had the data: registered vehicles, real-time range estimates, charger density maps, route planning. But accessing any of it meant navigating menus. The vision was a conversational assistant: ask it where to charge, how far your car will get on 50% battery, or what to do when a charging session gets stuck.
Dual-Mode Agent on Bedrock AgentCore
A single agent architecture handling both FAQ retrieval and live API task execution. Strands SDK handles function-calling, connecting the agent to EV.analytica's backend via JWT-authenticated API calls.
Cost-Effective RAG Without OpenSearch
FAQ content stored in S3 and indexed using Amazon Bedrock's native S3 Vector storage with Titan Text Embeddings V2.
Personalised Context from Live Data
Every session starts with user context pulled from RDS. When a user asks "how far can I get on 50% battery?", the response is specific to their car.
Guardrails and Agent Stability
System prompt rebuilt using a structured 4-pillar format based on Anthropic's best practices. Bedrock Guardrails deployed via Terraform.
In live testing, the agent returned accurate range estimates specific to the user's registered vehicle ~160km for a BMW i3 at 50% charge. EV.analytica now has a working conversational AI layer ready to sit on top of their platform. Next step: integrating the agent into the mobile app with streaming audio.
Tell us more about yourself and what you're got in mind.
Prefer email? hello@apexlab.io
Or even a quick call?
We don’t care about fancy CVs or long interview processes. Provide some basic information, let’s have a quick chat, and then show us what you love doing the most: write some code and create a product.
