hi, i'm tiffany! currently studying cs & math @ purdue, and spending this summer as a returning SDE intern at AWS. i'm drawn to problems involving systems, data, and decision-making under uncertainty, including backend infra, ml & agents, robotics, and quant.
outside of coding, you can usually find me climbing/hiking, playing poker, taking photos on my Sony a6700, or exploring new cafes!
Designed and implemented rules-based search indexing for a distributed document store built on DynamoDB and OpenSearch, enabling field-based routing rules that partition documents across dedicated indices to bound index sizes and improve query performance.
Worked across the full backend stack — from data modeling to the search API — and partnered with senior engineers to design and ship the feature safely to production.
Extended the search API with an optional rule-scoped query parameter for lower-latency targeted searches, while preserving full backward compatibility.
Expanded features of an AWS-native ticket and issue management API handling 130M+ daily requests, adding support for priority and label-based filtering to improve ticket triage and workflows.
Implemented end-to-end changes through model, service, and datastore layers — including enum validation, duplicate-label handling, and event auditing — and partnered with senior engineers on API schema decisions across 7 ticket operations.
Authored a technical design document, contributed to code reviews, and designed unit tests, integration tests, and canaries to validate API behavior across CRUD operations and failure cases for a safe two-phase production rollout.
Built an NLP-powered resume evaluation tool in Python that performs semantic analysis of keyword and action-verb usage to deliver actionable improvement recommendations to users.
Developed a Pandas preprocessing pipeline to clean and tokenize job postings for training an LLM job search copilot.
Project Lead — Social Stock Exchange @ Purdue Stack
Sept 2025 — Dec 2025
Led a 6-person team to ship a real-time celebrity news feed as a core component for a Social Stock Exchange platform.
Implemented a web scraping and LLM pipeline using Axios, Cheerio, NewsAPI, and Gemini to ingest celebrity-related news articles and developed a ranking algorithm to surface trending topics based on trend velocity, entity mention frequency, and recency.
Owned the technical architecture, project scoping, weekly sprint planning, and code reviews across the full stack.
Built vision-guided robotics system for autonomous drumming using Baxter robot with AprilTag-based localization and spatial tracking.
Processed camera input with OpenCV to detect AprilTags and compute 3D coordinates, translating targets into joint-space motions via inverse kinematics.
Enabled dynamic recalibration and precise rhythmic strikes by integrating perception, motion planning, and ROS-based control pipelines.
Engineered robot localization and path-planning using AprilTag fiducial markers, odometry, and motion profiling to enable autonomous navigation, auto-aiming, and real-time pose estimation during matches.
Spearheaded development of core robot subsystems including drivetrain, computer vision, and game-piece handling, applying PID and state-space control across autonomous and teleoperated operations.
Cleaned and analyzed Netflix dataset using Pandas and NumPy to identify trends across genres, release years, and ratings.
Built data visualizations with Matplotlib to highlight patterns in content distribution and viewer metrics.
Applied data preprocessing and exploratory analysis techniques to extract insights from large datasets.
Python · NumPy · Pandas · Matplotlib
projects.
🏆 Berkeley AI Hackathon — Top 10 Finalist out of 400 teams · Hackers' Choice Winner
Inspector
A multi-agent computer-use harness that lets coding agents QA their own work. an MCP server that plugs into Claude Code, Cursor, or any MCP client — it launches your app in a sandbox, splits the UI into regions, and runs one agent per region in parallel to screenshot, click, and verify. because it clicks by pixel instead of by selector, it works across web, Electron, Android, and native mobile.
Python · MCP · Claude · Computer Use
BoilerCheck
A RAG pipeline enabling source-backed Q&A across 450+ pages of Purdue University policy documents. Built with LangChain and Pinecone for retrieval and Gemini for answer generation. Benchmarked 30+ retrieval configurations and integrated reranking to improve MRR@4 by 14.3% and push hit rate to 100%.
Python · LangChain · Pinecone · Firebase · Gemini
Stock Market News Sentiment Analysis
A financial news sentiment analysis pipeline using FinBERT to score articles across 13 assets from Yahoo Finance, MarketWatch, and NewsAPI. Detects short-term shifts in stock market sentiment and generates hourly buy/sell/hold signals from sentiment deltas, visualized in a Streamlit dashboard.
Python · Pandas · PyTorch · Hugging Face · Streamlit
SkyLog
A full-stack flight tracking web app featuring an interactive 3D globe visualization of past flights and a reporting feature that aggregates flights by airline, route, and date to surface stats like total flight time, total miles, and on-time rate.
Python · Flask · PostgreSQL · React · Pandas
Functionary
An interactive tool for developers to visualize complex codebases, streamlining onboarding processes. Uses Abstract Syntax Tree parsing to map function calls and dependencies, with a RAG pipeline powered by Pinecone for semantic search and context-aware LLM chatbot queries.
A Raspberry Pi-based IoT system that streams camera data to a self-trained TensorFlow CNN for real-time wildlife detection, achieving 80%+ accuracy and triggering automated alerts for high-risk sightings. Includes a React Native mobile app with Firebase storage and Twilio SMS integration to deliver notifications. Built at the Management & Technology Summer Institute hosted by the University of Pennsylvania Jerome Fisher Program in Management & Technology.
i'm currently open to new opportunities in SWE, ML, quant trading, or any role where i get to work on cool, challenging, or impactful problems! whether you have a question, want to collaborate, or just want to say hi — feel free to reach out.