Hello! I'm David, a passionate programmer and full-stack engineer with leadership and research skills. I'm on a mission to make a significant impact in my community and beyond. I'm a former Google Software Engineering Intern where I worked on parsing code repositories for LLM context. Currently working on applying my skills in deep learning through research.
- INIT FIU - President - Miami, Florida (Jan 2024 - December 2024)
- Former Director of Technology (Jan 2023 - December 2023)
- Skills: Full-Stack Development, TypeScript, Docker, CSPs (AWS), Project Management
- Google - Software Engineering Intern (May 2024 - August 2024)
- Location: Sunnyvale, California
- Developed and designed end-to-end RAG system in Kotlin, Go, and TypeScript for user git repositories, informing external and internal chat systems on relevant code for user queries, integrating with larger micro-service architecture / frontend.
- Constructed pipeline for processing repositories, fetching repositories from distributed crawling service, building an Abstract Syntax Tree for finding important snippets, generating code-specific embeddings, and doing similarity search based on a user query.
- Prev. SWE + PM Intern @ Microsoft (May 2023 - August 2023)
- Location: Redmond, Washington
- Used SQL-like language for data analytics of big data to decide coverage of a Bing Explorational Search Box, settling on 1% of Bing traffic - 1 million daily users.
- Studied and implemented prompt engineering for the generation of content used in a Bing Explorational Search Box, then made change with C# framework so results can be passed to the frontend UX teams.
- Intro to Gradient Descent: A Python Notebook which goes through how gradient descent work and implementing a scalar auto-differentiation engine from scratch.
- Neural Network Architectures with Jax: Implementations of a linear regression model, MLP, CNN, RNN, and RNN w/ LSTM mechanism from scratch using Jax.
- GPT-2 From Scratch: From-scratch implementation of GPT-2 in Jax, trained on OpenWebText dataset and improved using a KV-Cache for faster inference & training.
- Intro to Ollama & HuggingFace: Begginner friendly guide on using Ollama for running models locally and HuggingFace for pipelines, inferencing, and uploading open-source models or datasets.
- Optimizers from Scratch: Implemented SGD, SGD w/ Momentum, Nesterov SGD, Adagrad, RMSProp, and Adam from scratch while testing against the baselines specified in the Adam: A Method for Stochastic Optimization paper.
Feel free to connect with me on LinkedIn or check out my website for more details about my projects and experiences.