Skip to content
View joaquinramirezg's full-sized avatar
6 followers · 5 following

Achievements

Achievement: YOLOAchievement: Pull Sharkx2Achievement: Arctic Code Vault Contributor

Achievements

Achievement: YOLOAchievement: Pull Sharkx2Achievement: Arctic Code Vault Contributor

Highlights

  • Pro

Block or report joaquinramirezg

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.

Content in all repositories owned by your account will be closed.
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
joaquinramirezg/README.md

Hi, I'm Joaquin Ramirez

Machine learning engineer and data scientist, now an MSBA candidate at UCLA Anderson ('27) in Los Angeles, headed for AI product management, product-focused data science, and forward deployed engineering.

Over the last three years I shipped production systems across HR technology, consulting, and consumer industries. Most recently I led development of Crehana's AI copilot and agentic platform (LangChain and LangGraph agents, RAG over a 1.2M+ vector corpus in Qdrant), used by 1300+ organizations and 48K+ users across Latin America. As a data scientist I built real-time fraud detection across a network of financial institutions, combinatorial optimization for contract allocation, computer vision for document intake, and retail sales forecasting.

That work is proprietary, so the case studies live on my portfolio:

→ joaquinramirez.dev · LinkedIn

Public here: GENIE, my undergraduate thesis on LoRA fine-tuning Falcon-7B for a Spanish-language university-regulations chatbot, and safe docs, a college platform where opening a shared document requires a live face check.

The rest are things I built to understand how they work, not to ship: a MIPS processor in Verilog assembled from CMOS gates up to a multicycle datapath, a CPU ray tracer written without a graphics API, and a graph library whose shortest-path algorithms route over a real airport network.

Pinned Loading

  1. genie-llm-chatbot genie-llm-chatbot Public

    Undergraduate thesis (UTEC): Falcon-7B LoRA fine-tune for a Spanish-language chatbot on university regulations. Eval loss -84%, perplexity 18.5 to 1.6.

    Jupyter Notebook 1

  2. safe-docs safe-docs Public

    Safe docs: college web platform where opening a shared document requires a live webcam face check (Flask, OpenCV, dlib).

    JavaScript

  3. joaquinramirezg.github.io joaquinramirezg.github.io Public

    joaquinramirez.dev, my portfolio: case studies in GenAI agents and RAG, fraud detection, and optimization. Plain HTML, no frameworks.

    HTML 1

  4. graph-algorithms-cpp graph-algorithms-cpp Public

    Graph library in C++: Dijkstra, A*, Bellman-Ford, Floyd-Warshall, Kruskal, Prim, strongly connected components, BFS and DFS, with a JSON parser that loads a real airport network.

    C++

  5. mips-processor-verilog mips-processor-verilog Public

    A single-cycle and a multicycle MIPS processor in Verilog, built up from CMOS gates through flip-flops, state machines and an ALU. 131 modules, a testbench at every level, runs under Icarus Verilog.

    Verilog

  6. ray-tracing ray-tracing Public

    A CPU ray tracer in C++20: vector maths, ray-object intersection, mirrored surfaces and light shading, written without a graphics API.

    C++