Patrick Mendoza

Miami Beach, FL

About

Hi there, I'm Patrick. I'm a rising 3rd year at Berkeley studying EECS and Applied Math. Currently, my interests lie within the intersection of Computer Science and Math such as Software Engineering, Machine Learning, and Cybersecurity. In particular, I'm very interested in Adversarial Machine Learning. If I needed to describe myself in one word it would be tenacious.

Experience

Berkeley Artificial Intelligence Research (BAIR) Lab - Undergraduate Research Assistant

Python, PyTorch

• Working on NLP with the Berkeley Speech Group.

Stripe - Software Engineer Intern

Java, Scala, SQL, AWS Java SDK, DogStatsD, Apache Flink, Kafka, Spark

• Developed and deployed a Flink Application that identifes and auto-terminates degraded Hadoop nodes
• Reduced degraded node identification and termination for an on-call Engineer from 45 to 0 minutes, a 100% improvement in response time.
• Decreased the number of pages an on-call Engineer would receive for potentially degraded Hadoop nodes from 30 pages a month to 9 pages a month, a 70% reduction.
• Integrated the AWS Support API to automatically create support tickets for degraded nodes, once they were terminated, to ensure that no degraded node be recomissioned back to Stripe.
• Created an efficient and scalable heuristic algorithm to identify degraded Hadoop nodes with a 90% success rate.

RISING Lab, University of Florida - Undergraduate Research Assistant

Python, Pandas, NumPy

• Developed Connected and Autonomous Vehicle (CAV) trajectories in Reduced Speed Work Zones (RSWZ)
• Simulated and explored various attack methodologies in V2I/V2X communications for CAV trajectories
• Worked presented at the ACM MobiHoc 2023 Conference and published in the ACM Digital Library

ACE Lab, UC Berkeley - Undergraduate Research Assistant

Python, Pandas, NumPy, OS, PrairieLearn

• Developed a cheat-detection program using the five indicators of cheating for take-home exams for PrairieLearn
• Integrated various statistical inferential techniques and methods to assess the likelihood of a student cheating.
• Processed and analyzed 1,500+ take-home online exam records from PrairieLearn and Berkeley.
• Optimized the runtime of the program from 60s to 10s, an 83% reduction in runtime
• Work presented at the SIGCSE 2023 TS and published in the ACM Digital Library

The Beauty and Joy of Computing, UC Berkeley - Student Assistant

XML, HTML, Spanish translation, CS Education

• Actively communicating with 6+ developers and designers to create efficient, readable code and content for the BJC curriculum.
• Bolster the structural integrity of the website’s codebase for better readability for onboarding members.
• Automate the translation of the Beauty and Joy of Computing online curriculum into Spanish for 30,000+ students nationwide.

Projects

Quantitative Trading Strategy

Python, Pandas, NumPy, Selenium, Plotly, Yahoo Finance API

• Implemented a quant trading strategy rooted in VIX pricing, VIX futures contango, VolDex pricing, and the S&P 500 Developed 10 key metrics that would be converted into a weighted average, then transformed into a percentile rank to determine what strategy should be taken for a particular trading day
• Built a web scraper using Selenium to collect VIX Term Structure data (historical prices, VIX futures contract expiration dates, and VIX futures contango) since 2007
• Utilized Pandas and NumPy to integrate data from distinct sources characterized by their different formats

2D Tile Exploration Engine

Java

• Designed and implemented a 2D tile-based world exploration engine.
• Utilized the Gaussian pseudorandom number generator to create random yet predictable worlds based on the inputted seed.
• Added a loading and saving mechanism to the game by detailing a user’s current state in the world into a text file.

Style Transfer

Python, TensorFlow

• Implemented an image generator that creates a new image that has content from a “content” image with style from a “style” image
• Built a custom cost model for Neural Style Transfer for the VGG-19 network from scratch, and defined content and style layers

Certifications

DeepLearningAI - Deep Learning Specialization

• Learned and built from scratch the following neural network architectures: NNs, CNNs, RNNs, BNNs, LSTMs, GRUs, Transformers
• Optimized neural network architectures using Dropout, Batch Normalization, Xavier/He Initialization, and Hyperparameter Tuning
• Applied industry-level frameworks such as TensorFlow to speech recognition, machine translation, and NLP tasks and more

Contact

patmendoza6745@berkeley.edu