cv

Manufacturing Engineering Intern

New Hampshire Ball Bearings

• Reviewed and updated 300+ machine setup sheets for turning operations, incorporating current tooling, blueprint-based dimensions/tolerances, and material-specific parameters.
• Optimized feeds, speeds, and tool selections to improve production efficiency, reducing cycle times by 5%–30% per part while maintaining quality specifications.
• Collaborated with manufacturing engineers, line leads, and machine operators to resolve non-standard cases, validate changes, and standardize setup documentation across production lines.

Machine Learning Engineer

AI Student Collective

• Created an AI-powered e-commerce tool that generated optimized descriptions across 200+ products.
• Engineered text and price/meta features and solved severe class imbalance (900k rows; 12k positives) using hard-negative undersampling; validated with stratified splits.
• Built a Streamlit app for interactive inference with prompting, side-by-side diffs, and guardrails for factual and brand compliance.

Manufacturing Intern

New Hampshire Ball Bearings

• Authored 3 illustrated operator manuals for retainer-line machinery in 3 months, tripling historical documentation output.
• Collaborated directly with machine operators, line leads, and the manufacturing manager to capture tacit process knowledge and validate procedures.
• Standardized machine operation, tool change, and part-inspection workflows using visual documentation, reducing process variance and ambiguity on the production line.

Bachelor

University of California, Riverside

Data Science

MyDreamerV2

Dreamer-style reinforcement learning agent in PyTorch using an RSSM world model and actor–critic imagination rollouts.

• Reached 70% success after 280k environment steps.
• Added Plan2Explore-style exploration bonus and enabled training on Apple Silicon via MPS.
• Implemented episodic replay buffer and improved stability with gradient clipping, EMA target networks, and checkpoints.

SaleSense

Inference app and model pipeline for scoring product listings and generating improved descriptions.

• Optimized descriptions across 200+ products.
• Handled severe class imbalance (900k rows; 12k positives) with hard-negative undersampling.
• Streamlit interface with prompting, diffs, and compliance guardrails.

IntelliView

A real-time interview coaching tool that analyzes facial emotions and speaking dynamics to generate a feedback timeline and post-interview report.

• Built a real-time interview coaching tool that detects facial emotions and speaking dynamics, then generates a timeline and an interview report with feedback.
• Trained a custom YOLO model for face/region detection and integrated it with a PyTorch emotion classifier (75% test accuracy)

Technical Blog

Step-by-step posts explaining AI/ML papers with diagrams, examples, and references.

• Published on personal blog.
• Published articles on Medium with Humans For AI.