Trivan Menezes

Trivan Menezes

Princeton CS | Microsoft Azure AI

Building AI tools for search, user experience, and language.

I'm currently interested in building comprehensive search and retrieval systems, creating new interfaces for the next generation of AI applications, and tailoring modern AI models to low-resource languages.

My work at Micrsoft Azure AI focuses on building a platform for enterprise search and retrieval-augmented generation.

My past work involves Computational Linguistics and Knot Theory.

Moments

Launched collab with Vercel

2025

Created the official Azure AI x Vercel AI SDK RAG chat app template. Open-source.

Placed at X Developer Challenge

2024

Came in 3rd place at first-ever X hackathon. Created VLM-based natural-language feed filter. Presented to Elon Musk.

Published Math Research

2021

My knot theory research was published in the Journal of Knot Theory and Its Ramifications.

Princeton University

2019-2023

Studied Computer Science at Princeton University.

First Microsoft Internship

2021

Started my journey at Microsoft as an Explore Intern, working on the Azure Cloud Supply Chain.

Joined Azure AI

2023

Began my full-time career at Microsoft working on Azure AI services.

Projects

Fine-Tuning SLMs for Low-Resource Indonesian Languages

python
trl

Explored fine-tuning techniques on open-weight models for Javanese (80M+ speakers) translation. Tuned a 4B model to match performance of GPT-4o and Gemini 2.0 Flash, showing a resource-efficient path for adapting models to underserved linguistic communities.

NY Signal

next.js
azure

Built a live application to generate daily news updates for New Yorkers, tailored to their neighborhood and topic preferences. Users can share their digest with community members.

Portuguese Etymology Visualizer

python
flask
html/css/js

Intrigued by my study of linguistics and the Portuguese language, I sought to explore the different influences on Portuguese and show how words from different ancestor languages evolved to their present form.

Translating Indigenous Indonesian Languages

pytorch
python

Indonesia is both one of the most linguistically diverse and most populated regions of the world. I created a pipeline to process datasets for indigenous Indonesian languages and create a translation model in PyTorch using a transformer.

National Team Mapper

python
flask
html/css/js

Built a Python script which scrapes roster data off of Wikipedia and records the birthplaces of players on every national team, displaying this data on an interactive map.

TigerGig

python
flask
html/css/js

Built TigerGig, a creator marketplace for students at Princeton University to find talent to hire. Integrated OAuth security, designed an aesthetic UI, and wrote in-depth technical and user guides.

music.me

html/css/js
mathematica

Worked with three friends to build a web app which curates music playlist sequences depending on the user's schedule, using Google Calendar, Wolfram Mathematica for classification, and the Spotify API.

Knot Theory Research

mathematica

Worked with Dr. Claus Ernst at Western Kentucky University on a project in the theoretical math field of knot theory, resulting in a published research paper.

Experience

PresentSoftware Engineer 2Microsoft Azure AI Search
2023Software EngineerMicrosoft Azure AI Language
2022Software Engineer InternMicrosoft Azure AI Vision
2021Explore InternMicrosoft
2019Software Developer InternInterapt
2018Undergraduate Researcher (Knot Theory)Western Kentucky University