Alex

• Develop custom virtual formula field functionality in Salesforce Data Cloud, empowering customers to create dynamic data transformations without code and enabling self-service analytics capabilities.
• Build scalable backend services using Java and Apache Spark for formula computations on distributed metadata services, enabling processing that scales automatically with customer data volume.
• Implement REST API endpoints with robust error handling and validation, providing developers with reliable programmatic access to formula field operations and reducing integration complexity.
• Create intuitive React and TypeScript frontend interface that simplifies virtual formula field creation and data exploration, dramatically improving user experience and accelerating customer onboarding.

• Maintain Cornell's course management system, used by 10,000+ students in 100,000+ line codebase.
• Lead frontend team by designing project architectures, reviewing PRs, and interviewing candidates.
• Architected and built the student-facing frontend in React/TS, implementing a component architecture and state management to reduce loading times by 90% from the previous implementation.
• Implemented APIs with serializable data structures, enabling secure communication between frontend and backend while improving developer velocity through enhanced code modularity and reusability.

• Developed AI agent for QuickFi, a commercial lending technology company, automating insurance certificate validation and compliance verification processes, reducing operational costs.
• Built frontend and vector database for ClassGPT, enabling professors to create AI tutors for classes.

• Worked on the interpreter for SASyLF, a language for writing proofs, implementing core features in Java.
• Implemented a polymorphic type system for SASyLF, similar to Java generics, enabling type-safe code reuse and significantly reducing duplication in proofs by allowing abstraction over multiple data types.
• Created a modular testing framework using Python to automate integration tests, ensuring reliability.

High-performance math library supporting numerical methods, calculus, and linear algebra. Includes 15+ ML models and RL agents.

• Developed a high-performance math library supporting numerical methods, calculus, and linear algebra.
• Implemented 15+ ML models, such as neural networks, achieving 99%+ accuracy on classification tasks.
• Implemented DQN and Policy Gradient agents achieving 97%+ win rates in autonomous gameplay.
• Achieved 26x performance speedup through compiler optimizations and parallelization techniques.

Custom functional programming language with a performant interpreter, type inference, and core language features.

• Built interpreter for custom programming language, achieving 50% speed improvement over Python.
• Designed parser and type inference algorithm to validate programs and provide developer feedback.
• Implemented core language features including pattern matching, functions, and polymorphism.

Full-stack habit tracking application with OAuth2, monthly calendar interface, and real-time goal management.

• Built full-stack habit tracking application with OAuth2, calendar interface, and goal management.
• Developed TypeScript APIs and Express.js backend integrated with Firebase for seamless persistence.
• Implemented live chat functionality using database connections, enabling low-latency communication.

Evolving artificial life simulator with JavaFX GUI and custom programming language for organism behavior.

• Built JavaFX GUI app to visualize real-time interactions and evolution of virtual organisms.
• Implemented custom programming language with parser and interpreter to model organism behavior.
• Developed pathfinding algorithms including A* and Dijkstra's for organism navigation.

Show, Attend, Tell implementation: attention-based image captioning model with superior METEOR scores.

• Implemented attention-based image captioning model with superior METEOR scores vs. original paper.
• Built soft and hard attention mechanisms with REINFORCE and backpropagation for training.
• Developed encoder-decoder architecture using ResNet-50 and LSTM with attention visualization.

Interactive platform for visualizing and experimenting with classic algorithms and data structures. Designed for students, educators, and developers to learn and teach algorithms in an engaging way.

• Built interactive platform for visualizing classic algorithms and data structures with real-time animations.

AI-powered assistant for Cornell students that matches users to courses and professors based on their academic interests, backgrounds, and goals. Aggregates course and professor data for quick, informed decisions.

• Matches students to courses and professors based on academic interests and goals.
• Aggregates course and professor data for quick, informed decisions.
• Uses Pinecone, a vector database, for fast and intelligent course matching.
• Features a modern, user-friendly frontend for a seamless experience.
• Database includes 19k courses and 8k faculty.

This portfolio website