Skip to content
View cauafsantosdev's full-sized avatar
👀
👀

Highlights

  • Pro

Block or report cauafsantosdev

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
cauafsantosdev/README.md

Hi, I'm Cauã Santos

Machine Learning / MLOps Engineer | Backend & Cloud

I'm an Information Systems undergraduate at the Federal University of Rio Grande (FURG), graduating in 2027, and currently working on applied R&D projects at iTec/FURG's EMBRAPII unit.

I focus on the engineering behind machine learning systems, from models and data pipelines to backend services, cloud infrastructure and deployment.

My current work spans optimization, ML systems, AWS infrastructure, LLM applications and recommender systems, primarily using Python, Go and Terraform.

Selected Projects

NexdWatch | Recommendation Engine

End-to-end movie recommendation system built over 4.3M+ Letterboxd interactions.

Implemented hybrid SVD/FAISS + popularity candidate retrieval, reaching 84.3% candidate recall at a 2,000 item retrieval budget and 99.8% catalog coverage in offline evaluation. The system also includes asynchronous data ingestion, reproducible offline ML pipelines and a FastAPI inference service.

Python · FastAPI · PostgreSQL · FAISS · scikit-learn · Docker

Lexos | Event Driven AI Document Processing

Self hosted AI document processing platform built around an event driven architecture and designed to run efficiently on CPU only infrastructure.

A Go API gateway coordinates asynchronous Python workers for document summarization, RAG and speech transcription. The system includes Redis backed task state, content addressed processing, S3 compatible object storage, SSE streaming, artifact caching and local LLM inference with Qwen.

Live application →

Go · Python · Redis · Qwen · llama.cpp · FAISS · Faster Whisper · Cloudflare R2 · Next.js

YADN | Serverless AI Newsletter

Fully automated serverless newsletter that collects and curates content from 14 AI and software engineering sources every weekday.

A Go pipeline running on AWS Lambda processes the feeds, uses Amazon Bedrock to evaluate and summarize selected articles, persists editions in DynamoDB and distributes the newsletter through SES, at approximately $2/month in LLM inference costs.

Live dashboard →

Go · AWS Lambda · EventBridge · Amazon Bedrock · DynamoDB · SES · S3 · CloudFront · AWS SAM

vps-infra | Self Hosted Infrastructure

Reproducible shared infrastructure for running multiple portfolio applications on a single VPS without turning the server into an unnecessarily complex platform.

Provides reverse proxying and automatic HTTPS, shared PostgreSQL and Redis services, Prometheus and Grafana observability, container and host metrics, automated PostgreSQL backups to external object storage, Docker network isolation and Debian bootstrap automation.

Docker · Caddy · PostgreSQL · Redis · Prometheus · Grafana · Linux · Shell

Current Work

At iTec/FURG, I work on two applied R&D projects.

CELIA focuses on backend engineering and Genetic Algorithm based fleet optimization, including multi objective fitness functions, operational constraints and scenario evaluation.

LAPIS focuses on Machine Learning and AWS infrastructure engineering, with reusable Terraform modules, container orchestration, serverless inference and automated cloud deployments. I authored core Terraform infrastructure for ECS, RDS, Lambda and S3, as well as most of the orchestration for its main AWS deployment scenarios. The most complex environment can be provisioned completely with a single terraform apply in approximately 10 minutes.

Previously, during the BRISA ICT Residency, I designed and implemented the complete backend of an administrative platform that was deployed and used by a psychology clinic.

Tech Stack

Machine Learning & AI
PyTorch · scikit-learn · LightGBM · FAISS · NumPy · Pandas · Recommender Systems · LLMs · RAG · Computer Vision · Genetic Algorithms

Cloud & MLOps
AWS · Terraform · Docker · Git · GitHub Actions · CI/CD · Linux

Backend & Data
Python · Go · TypeScript · FastAPI · Django REST Framework · Echo · PostgreSQL · Redis · SQLAlchemy · Alembic

Architecture & Systems
Event Driven Systems · Asynchronous Processing · Serverless · Infrastructure as Code · REST APIs

About Me

I'm particularly interested in problems at the intersection of Machine Learning, Software Engineering and Infrastructure, especially ML Engineering, MLOps and AI Engineering.

Based in Rio Grande, Brazil, and open to remote internship and junior opportunities.

Outside engineering, I'm into cinema, chess and music.

Contact

LinkedIn · Email

Pinned Loading

  1. nexdwatch nexdwatch Public

    Personalized Letterboxd recommendation engine with zero-shot SVD profiles, FAISS retrieval, categorized feeds, async ingestion, and a versioned MLOps pipeline built with FastAPI, Next.js, Celery, P…

    Python

  2. lexos lexos Public

    Event-driven AI document processing engine with offline RAG, summarization, speech transcription, SSE streaming, and self-hosted LLM inference.

    Go 5

  3. yadn yadn Public

    A fully serverless AI newsletter and web dashboard for software engineers. Powered by a Go Lambda scraper, DynamoDB deduplication, LLM context-window optimization via Amazon Bedrock, and deployed g…

    Go

  4. vps-infra vps-infra Public

    Shared Docker-based VPS infrastructure with PostgreSQL, Redis, Caddy, Prometheus, Grafana, backups, and Debian bootstrap automation.

    Shell