Blog – Valeria Micol Garcia – Computer Engineering Student at UNLP | AI & Machine Learning Enthusiast

Blog

Technical notes by domain — for reviewing concepts

Deep Learning & AI

Neural architectures (Transformers, CNNs, LSTMs) and multimodal models (e.g. Grounding DINO).

LangChain: theory and implementation in a nutritional label analyzer

LangChain LangGraph LLM Gemini Tavily

What LangChain is (models, messages, tools), how it relates to LangGraph, and how it is used inside the nodes of a nutritional label analyzer microservice (Gemini, Tavily).

LangGraph: theory and implementation in a nutritional label analyzer

LangGraph LangChain Agents LLM Gemini

What LangGraph is (state graphs, nodes, edges), how it relates to LangChain, and how it was implemented in a microservice that analyzes nutritional labels with Gemini and Tavily.

How to add a RAG to your project — step-by-step with LlamaIndex, FastAPI, and Groq

RAG LlamaIndex FastAPI Groq Nutri-AI

Simplified tutorial to add a RAG (Retrieval-Augmented Generation) to a project: folder structure, dependencies, env vars, LlamaIndex engine, FastAPI endpoint, and how it was done in Nutri-AI.

What is LlamaIndex?

LlamaIndex RAG LLM Nutri-AI Groq

LlamaIndex is an open-source framework for building apps that connect your data (documents, PDFs, databases) with LLMs. It orchestrates indexing, semantic search, and context for RAG. How it is used in Nutri-AI.

What is RAG and how it is implemented in Nutri-AI

RAG LLM Embeddings Nutri-AI Groq

What an LLM is, what RAG (Retrieval-Augmented Generation) is, why embeddings are used, and how RAG is implemented in the Nutri-AI project: indexing, chat flow, and models used.

Grounding DINO — Model Overview and Concepts

Object Detection Transformers Vision-Language Hugging Face

How Grounding DINO is used in the Zero-Shot Object Detection Service: where it comes from, how it is invoked in the code, and how it was put into practice (API, deployment).

PyTorch vs TensorFlow — main differences

PyTorch TensorFlow Deep Learning Neural Networks Machine Learning

Key differences between PyTorch and TensorFlow: execution model, API style, deployment, ecosystem, and when to choose each for deep learning.

MLOps Integration in the Zero-Shot Object Detection Service

MLOps Human-in-the-Loop Grounding DINO Zero-Shot Detection Supabase

How a human-in-the-loop (HITL) flow was added to collect user corrections on ingredient detection (Grounding DINO), and which MLOps concepts are applied — data storage, annotation format, governance, and future evaluation/training pipelines.

Transformers — What they are and main architectures

Transformers BERT GPT Vision-Language Grounding DINO

A short overview of what a Transformer is and the main types of transformer architectures (encoder-only, decoder-only, encoder-decoder, vision, vision-language), with examples and typical use.

Machine Learning & Statistical Modeling

Tabular and biomedical modeling: Random Forest, XGBoost, regressions.

White Box vs. Black Box Models: The Trade-off Between Explainability and Performance

Interpretability Random Forest Research Black Box White Box

In biomedical and clinical research, interpretability matters: black box vs white box models, why Random Forest and similar methods are often preferred over deep learning, and how this applies to lab work with physiological signals.

Software & Backend Engineering

Design patterns, Pydantic, dependency injection, microservices in Python.

Pydantic: theory and implementation

Pydantic FastAPI Validation Python Nutritional Labels

What Pydantic is (BaseModel, Field, validation, serialization), why it is used in APIs, and how it is used in the nutritional label analyzer, the RAG service, and the Nutri-AI ingredient-detection backend.

Deploying a model on Hugging Face Spaces

Hugging Face Spaces Docker Deploy Nutri-AI

How to deploy an ML app (e.g. FastAPI + Grounding DINO) on Hugging Face Spaces: what Spaces is, SDK options, concrete steps, and how Nutri-AI meets the requirements.

Docker — Overview, basic implementation, and use in Nutri-AI

Docker Containers FastAPI Nutri-AI

What Docker is, how it is implemented at a basic level (Dockerfile, build, run, .dockerignore), and how it was used in the Nutri-AI food-detection (ingredient-detection) backend.

Python (FastAPI) vs Node.js (Express) — backend comparison

Python FastAPI Node.js Express Backend

Key differences between building a backend in Python with FastAPI and in Node.js with Express: runtime, server model, validation, documentation, and when to choose each.

Node.js backend with Express — the basics

Node.js Express Backend JavaScript REST API

What you need to know to build a backend in Node.js with Express: server, routes, middleware, request/response, and how it compares to other backends. I learned this in the Backend Programming courses on my career timeline.

FastAPI — Python backends and the Nutri-AI ingredient-detection API

FastAPI Python Backend Pydantic Nutri-AI

What FastAPI is, what a Python backend is, and how FastAPI was used in the Nutri-AI food-detection API: app setup, CORS, Pydantic, routes, and documentation.

Data Engineering & Signal Processing

Wearable sensor signal processing, ETL pipelines, Pandas/NumPy.

No posts yet in this domain.

Systems & CS Fundamentals

Memory management, complex data structures, algorithmic optimization (C/Python).

No posts yet in this domain.