Lisa AI: From WhatsApp reminders to your personal JARVIS

LISA AI: Building a real-life JARVIS

A journey towards creating a digital assistant that understands and anticipates your needs.

Imagine for a moment that you could have a natural conversation with your house. I'm not talking about those frustrating moments when you yell at Alexa because you don't have Spotify Premium and it plays some random playlist. I'm talking about a real conversation with a system that doesn't just execute commands, but actually knows you, anticipates your needs and, by the way, remembers you're vegan without you having to repeat it every time you search for recipes.

That's Lisa, the project that started as a simple WhatsApp reminder system and evolved to become the definitive answer to the question: "Why can't my house be as smart as my smartphone?"

Plot Twist: From Reminders to Revolution

It all started as a personal experiment. I was building from scratch an expense and activity reminder system for WhatsApp using vanilla TypeScript. I know what you're thinking: "similar solutions already exist." But I was having fun, creating APIs manually, handling webhooks like a digital craftsman, when I realized something: I was reinventing the wheel, but with extra features.

And that's how the idea for Lisa was born (at that time it was called VACHO) - an artificial intelligence agent that isn't content with being just another domestic chatbot. It's the assistant you always wanted but the tech industry forgot to make truly intelligent.

What is Lisa AI really?

It's what you get when you combine the ambition to create a real JARVIS with a developer's obsession to do things right. It's a complete ecosystem that integrates:

Software: The Brain that Never Rests

LISA's core runs on state-of-the-art language models (currently Gemini, but with support for local models because yes, your privacy matters). But here comes the interesting part: it uses Model Context Protocol (MCP) - think of it as USB-C for AI applications.

What does this mean in practical terms? That LISA can connect to virtually any tool in your digital ecosystem:

Delivery Apps: Imagine ordering your favorite food without having to open an app. Simply tell LISA what you want and she'll take care of the rest.
Email: goodbye to the overwhelming inbox, you'll have an assistant that organizes your emails, responds to the most urgent ones and reminds you of what really matters.
Calendars, reminders, even WhatsApp.

Basically anything that has an API or can be automated. It's super glue for your digital life.

But LISA isn't just another command executor. It has full conversational mode: remembers anecdotes, offers advice, contextualizes, dialogues about personal topics and adapts to your profile. Are you vegan? It will suggest plant-based recipes. Need medication reminders? It will handle them with the discretion of a doctor.

Hardware: Where Magic Becomes Tangible

LISA's physical ecosystem is deployed through two devices that are basically the Avengers of the smart home:

The Main Gadget: Your New Best Friend

A speaker/microphone device with camera and interactive LEDs. Think Alexa, but with real personality and without that constant feeling that Amazon is listening to your intimate conversations.

Features that matter:

Natural voice interaction and multimodal capability
Contextualized visual expressions (LEDs don't just blink, they also communicate states and emotions)
Ability to take family photos and create albums automatically
Responses that actually make sense and feel human

SmartLink: The Normal House Transformer

This is where things get really interesting. SmartLink is an intelligent relay with microcontroller that converts any conventional device into smart. No batteries, no complications, just direct connection to your electrical infrastructure.

This means you can automate:

Lights (obvious, but necessary)
Legacy electronic equipment
Entertainment systems
Any device that uses electricity

All based on learned routines and real behavior patterns, not tedious manual programming.

A brutally honest comparison

LISA AI vs. Alexa: David vs. Goliath

Feature	Alexa	Lisa
Natural conversation	"Alexa, play music"	"LISA, I need a song that helps me concentrate"
Contextual memory	Forgets your preferences	Remembers you don't like broccoli from the shopping list you shared a month ago
Privacy	Everything goes to Amazon	Local models available
Real integration	Closed ecosystem	MCP = connects with EVERYTHING
Personalization	Limited skills	Digital chameleon

Rabbit R1 and Humane AI Pin

While Rabbit R1 bets on being your portable companion and Humane AI Pin tries to revolutionize wearable interaction, LISA takes a different approach: dominate your space, not compete for space in your pocket.

The best smart device to carry has already been created - the iPhone came out in 2008, but what about the home?

Rabbit R1: Great to carry, LISA is great to live with
Humane AI Pin: Wearable innovation, LISA is livable innovation

With Lisa, your home is the smart device

Stats For Nerds: The numbers that matter

Technical data is the real benchmark for evaluating development.

Architecture and Performance

Average response latency: 842 ms (includes full context processing)
Contextual memory: Scalable through ChromaDB usage, a user's complete context can be stored in approximately 700 MB.
Intent recognition accuracy: 92.4% in Spanish, 96.7% in English.

Stack used

code

fastapi + uvicorn # Robust and fast API
langgraph + langgraph-api # The brain of the operation
langchain-google-genai # Connecting with Gemini for tests
mem0ai # Intelligent persistent memory
chromadb # Vector database for semantic context
pydantic # Data validation that doesn't lie

Development metrics

Core LOC: ~11,500 (not counting tests, because we're professionals)
Test coverage: 82% (the remaining 18% are "features", not bugs)
Average response time to complex queries: 1,183 s
Available MCP integrations: 17 (and counting)

Current status: Where are we?

Current phase: Advanced software development with solid LangGraph architecture and hardware planning in progress.

Software: 72% completed

The core "skeleton" is practically finished. When I say skeleton, I mean all the infrastructure that will allow LISA to be truly intelligent, not just a chatbot with ambitions.

Completed:

Multimodal contextual conversation system
Integration with any language model
Functional MCP framework
Persistent memory system (both session and long-term)
Nearly complete API but documentation is missing

In development:

Personality fine-tuning
Contextual response optimization
Testing in real scenarios

Hardware: In active planning

This is where I admit I need help. Between university, work and other parallel projects, hardware needs a dedicated partner.

SmartLink is still in planning phase (the name is provisional, I accept suggestions that don't include "Smart").

The roadmap: Peru Tech Week

Immediate objective: Present LISA AI at Peru Tech Week (September 29 - October 5), specifically October 3-5.

It won't be a commercial launch, but a proof of concept demonstration that shows the real potential of having an assistant that truly understands and adapts to you.

What to expect in the Demo?

Real-time natural conversation
Basic home task automation
Integration with real services
Demonstration of contextual memory and personalization
SmartLink concept preview (possibly even a prototype)

The Vision: Beyond the Smart Home

LISA AI is not just another home automation product. It's the foundation for a marketplace of personalized integrations where the community can create, share and monetize specific connections for unique use cases.

Imagine:

Integration with your irrigation system that considers local weather
Connection with your fitness platform that adjusts temperature according to your exercise routine
Synchronization with your remote work that optimizes your home office space

The goal is to create an ecosystem where your home is not only smart, but wise.

Partner Search: Is it you?

I'm currently looking for a hardware partner who shares the vision of creating something truly revolutionary. If you have experience in:

IoT device development
Consumer electronics
Microcontrollers and embedded systems
Rapid prototyping and hardware design
Desire to innovate and create something unique

Contact me: rodrigozuniga3010@gmail.com

Final reflection: Why LISA AI Matters

In a world where AI is everywhere but rarely truly useful, LISA AI represents a return to fundamentals: technology that genuinely improves daily life.

It's not about impressing with technical demos or generating artificial hype. It's about creating an assistant that's as natural to interact with as a friend who happens to know how to connect to the internet and control your house.

Because at the end of the day, the best technology is the one that disappears, the one that works so well you forget it's there, until you need it.

LISA AI is building that future, one smart home at a time.

Questions? Comments? Recommendations? Collaborations?

Next stop: Peru Tech Week (September 29 - October 5) where I hope to participate and for LISA AI to finally come to life.

Stay tuned! This is just the beginning.