Alexa is Multimodal agent (Text, Image, Audio)
1.Autonomy: Functions as a simulated ML Engineer with a defined weekly schedule (Monday to Sunday), allowing for advanced task scheduling and state transitions.

2.Personification: Possesses short-term memory (context) and long-term memory
(knowledge), enabling a deep, ‘HER-like’ conversational experience.

3.Advanced Modality: Seamlessly processes and responds across Text, Audio, and
Image inputs and outputs

Alexa’s brain have two memories Short term and Long term
Short term memory is for conversation and long term memory to remember useful information which is in Qdrant (Vector DB).

for agent orchestration I have used Langgraph with SQLite for short term memory.

Newton LLM Project inspired from celebrity twin project, bring the legend of physics back to life with AI. There are 3 main component of this project

1. Data Ingestion Pipeline:
   Apache Airflow is used as data orchestrator in data ingestion pipeline. ETL flow, Extract data from Wikipedia and other URLs, transform it in structure manner, store it in MongoDB with _id, title, content, timestamp, source.

2. RAG Pipeline:
   Retrieval part is done from getting data from MongoDB make vector store using Qdrant, Generation is done via OpenAI GPT-4o-Mini model. Transform text into chunk and embed with OpenAI Embedding Model. Evaluation done based on Answer generation.

3. Deployment:
   Useful endpoints for getting Vector store, Chat, Health as RESTFul API using FastAPI and deployed on Render.