
Oracle
Fully local cited Q&A over Obsidian notes.
0.90held-out top-3 search score with hybrid search, vs 0.80 vector-only

I'm a software engineer at Intuitive Machines, where I build an agentic AI desktop application on Claude for aerospace engineering teams, and I'm working toward an M.S. in Artificial Intelligence at San José State. I'm drawn to LLM systems that show their work: retrieval that cites its sources, agents that use real tools, and evaluations that check whether any of it actually helps. Before AI, I studied computer science and engineering at UC Irvine, building compilers, processors, and embedded systems.
– Present
Supports the company's AI initiatives and enterprise software applications.
–
Lanteris was acquired by Intuitive Machines, and the role continued there.
–
LLM output evaluation.
–
Contributed to the team's software development.
Expected
Relevant coursework: Machine Learning · Deep Learning · Natural Language Processing · Cloud Technologies · Enterprise Software · Data Mining
Projects: Logbook · Autonomous Vehicle Cloud Platform
GPA 3.8 · Cum Laude
Relevant coursework: Compilers · Operating Systems · Computer Architecture

Fully local cited Q&A over Obsidian notes.
0.90held-out top-3 search score with hybrid search, vs 0.80 vector-only

Grounded Q&A over 3,080 NASA aviation maintenance safety reports, with every cited report ID checked.
22/22valid filter tool calls from Claude in the model screening
Turns course files (PDF, PPTX, DOCX) into practice questions, each citing an exact quote that code checks is on the cited page.
37/37benchmark facts recovered with Claude vision, vs 10/37 from text alone

Mastermind in the browser, rewritten in TypeScript from a Python terminal game, with four difficulty levels and two-player pass-and-play.

Multi-tenant cloud platform for simulated autonomous vehicles, from CARLA telemetry to hazard alerts on a fleet dashboard.
A few things I'm into outside work. Click a pin to check it out.