Software Engineer

• Extended Unity Catalog data lineage from tables to Volumes, widening the core abstraction from a table-centric model to a generic typed node graph across the event schema, the ingestion consumers, and both storage backends.
• Landed the change as a behavior-identical no-op under a zero-diff deploy rule, building the file path and volume layers incrementally behind feature flags, with a dual-write path so a single event produces both a legacy string-keyed row and a UUID-identified graph node, handling volume-to-path fan-out and deduplication at ingestion.
• Moved volume validity enforcement out of consumer-side inference and into construction time, establishing a single source of truth and reducing the cost of adding future securable types.

• Built a high-performance video overlay redirection system for the DCV Web Client, supporting multi-monitor setups with 120 FPS rendering and 0.15 ms latency.
• Developed a fault-tolerant communication layer between Chrome extension, proxy, and client to keep overlays stable during tab detachment and multi-display movement.

• Built a MuseScore plugin that integrated transformer-based music generation, enabling AI-assisted composition with real-time melody generation, style transfer, and sequence prediction.
• Developed large-scale data pipelines and visualization tools, working with 550k+ MIDI files, Python, SQL, and transformer models to evaluate accuracy, latency, and output diversity.

• Built and shipped a Chrome extension that automated LinkedIn post and comment engagement with Gemini, cutting manual outreach time and boosting reply rates.
• Developed full-stack integrations across LinkedIn, Gemini, and AWS Amplify APIs, using Node.js, Express, and React/TypeScript to support advanced features like personalized comments and history tracking.

Developed a portfolio optimization framework using Hidden Markov Models to detect latent market regimes from engineered financial features. Built a walk-forward machine learning pipeline with PCA-based dimensionality reduction and regime-aware mean-variance optimization to dynamically rebalance ETF portfolios. Evaluated over 84,000+ strategy configurations across a 14-year out-of-sample backtest, validating improved risk-adjusted performance and reduced downside risk through robust regime-aware asset allocation.

Received funding to implement this strategy from Georgia Tech’s Investments Decision Group.

A statistical arbitrage project using volatility skew and VIX term structure to forecast SPY returns. Combined rule-based signals with logistic regression in a volatility-targeted framework.

A quantitative research project analyzing whether overnight price gaps predict next-day realized volatility in SPY. Includes 20 years of data, volatility estimators, OLS regression, GARCH modeling, and rolling correlation analysis.

Machine learning project predicting chart longevity of Spotify Top 50 songs using audio features and YouTube engagement data. Built preprocessing pipelines, engineered collaboration and virality features, and trained models including XGBoost.

A decentralized Ethereum contract for milestone-based payments between clients and freelancers. Includes trustless escrow deposits, IPFS-linked milestones, and mediator-based dispute resolution.

A Game Boy Advance adventure written in C where players guide their soul through waves of fire to reach salvation. Features dynamic fireball patterns, karma-based scoring, and an unlockable Impossible Mode.

An Android app that helps people manage meals, recipes, and pantry items with real-time syncing and a clean, interactive design.

• Built a low-latency AI lecture copilot that streams live microphone audio for real-time transcription and in-class LLM Q&A, generating personalized notes, quizzes, and instructor insights.
• Engineered a full-stack system (React, Node.js, Convex, LiveKit, Gemini) to process long lecture sessions, detect confusion spikes, extract key concepts, and surface engagement trends.
• Designed a transcript chunking + compression pipeline reducing prompt token usage ~70% while preserving real-time responsiveness for extended sessions.

• Engineered a full-stack React + Spring Boot platform for International Girls Academy, transforming their outreach with scalable program management and high-engagement learning tools.
• Built secure auth, role-based dashboards, personalized course progression, and interactive modules that significantly improved usability for students and staff.

• Tutored undergrads in CS and Math courses including DSA, Linear Algebra, and OOP.
• Completed 300+ hours of one-on-one sessions with 4.98/5 rating.

• Manage six-figure budgets, dues, and tax filings for 100+ members.
• Develop policies to optimize cash flow and sustain chapter growth.

• Conducted hippocampal subfield analysis on MRI scans of neurotypical and Down syndrome youth using ITK-Snap and R.
• Contributed to a forthcoming publication and presented results at the AZBio Awards, gaining experience in neuroimaging, data analysis, and scientific communication.

• Assisted weekly classes and workshops in piano and music theory.
• Helped prepare students for grade exams through guided practice and group sessions.

• Supported instructors by grading assignments and organizing classroom activities.
• Reinforced lessons in math and English while tracking student progress.