Case Study
J.A.R.V.I.S.
Mark I
An AI assistant that converts natural, unstructured language into organized calendar events and tasks, powered by dynamic model routing (Ollama, OpenAI, Anthropic).
Problem
Managing time usually requires tedious clicking through calendars and note apps. Quick thoughts ("dinner with Anna on Friday evening") slip away if not cataloged immediately.
Solution
I built a backend assistant in Python:
- NLP to JSON — extraction of time, place, and intent from conversational speech
- Model Routing — intelligent model selection (from local Llama3 to GPT-4) based on task complexity
- Contextual Memory — understanding relative terms like "next week" or "the day after tomorrow"
Approach & Takeaways
This project was my entry into prompt engineering and building AI orchestrators. It taught me how to create systems where the software decides which LLM (Cloud vs. Local) to use to maintain the best balance between cost, speed, and precision.
PythonLangChain / LLM APIsOllamaData Extraction
View Source