Leon Do
・No songs playing at the moment
I'm Leon, a 22-year-old aspiring software engineer at the California State University of Long Beach, majoring in Computer Science with a minor in Pure Mathematics. My journey into programming began in high school around 5 years ago, sparked by my interest in the sneaker reselling industry. While building my first major project and sneaker bot, Rias Software, I realized I became obsessed with automation, reverse-engineering, and optimization.
Since then, I've completely fallen in love with building through the art of programming. For me, programming isn't just code. It's problem solving, system elegance under pressure, and the thrill of turning abstract ideas into performant, real-world solutions.
These days, I spend my time building & optimizing systems / infrastructure software, along with solving algorithmic challenges. Do please check out my blog where I write about my projects, thoughts, & random blurbs about things I find interesting. For my algorithmically hungry friends, I write detailed breakdowns on various competitive programming topics.
Recent Interests
Korean BBQ
Competitive Programming
C++
Curry Udon
Machine Learning
Operating Systems
Hardware Engineering
Cooking
Judo
Hyperpop
Experience
Zwei Labs・Software Engineering Intern
10/2025 - 1/2026
Developed a real-time Python trading backend for market data, live order books, strategy execution, and pre-trade risk checks. Scaled to 40+ concurrent markets using independent async workers and WebSocket streams, isolating slow feeds from unrelated markets. Added bounded queues, backpressure, deterministic event ordering, and exact decimal pricing to preserve correctness under load.
Beach Investment Group・Quantitative Developer Intern
6/2024 - 5/2025
Created a scalable Python/Polars data pipeline used by 25+ analysts for ingestion, preprocessing, model training, and prediction delivery. Accelerated NumPy with CuPy / CUDA, cutting experiment times from hours to minutes and increasing research throughput. Implemented reusable factor-model and time-series tools with scikit-learn; forecasts supported funding proposals that secured $115K.
Projects & Ventures
5/2026 - Present
Built a full-stack exam prep platform with React, TypeScript, Supabase, and Vercel that generates practice questions from user-uploaded material, used by peers to study for exams and technical interviews. Implemented autosave so users never lose progress mid-exam, syncing drafts to the cloud in the background with conflict resolution across devices. Built a one-click course sharing system where users generate a link and recipients instantly import the full course, ready to take exams with no setup.
Video Streaming Platform
5/2026 - 6/2026
Assisted in stabilizing a short-form video platform after a 20,000+ user launch spike by diagnosing performance issues and infrastructure bottlenecks. Reduced server load by 98% during a 20,000+ user launch spike by eliminating duplicate requests and preventing redundant API calls. Improved user load time from 14s to 1.3s by lazy loading chunked data instead of blocking on full upfront fetches.
MaguroPlace - GPU-Accelerated Placement Optimizer
Created a GPU-accelerated optimization engine in Python/PyTorch that produced valid layouts across all 17 benchmarks. Improved evaluation throughput 18x, from 11 minutes to 38 seconds, by profiling bottlenecks and caching reusable benchmark data. Reduced the mean challenge score 47% (2.09 to 1.10), beating the published 1.46 baseline.
ASIC Reverse Engineering
Built Python EDA tooling that automatically recovered a gate-level netlist and signal connectivity from a final GDSII physical layout. Generated structural Verilog and verified the recovered design through simulation against standard-cell models. Converted the circuit into a graph, collapsed feedback loops, and used 2-SAT to avoid repeated compile-and-simulate search loops.
DeepMew - Resumable Cloud Jobs
Ran distributed PyTorch workloads on Modal cloud workers with checkpointing, restart recovery, and cost controls for fault tolerance. Improved reliability by fixing a checkpoint bug that erased progress after worker restarts; verified the fix across more than 400K games. Scaled a rate-limited data pipeline from 1,543 replays to 180K training rows while staying within API limits.
Blotto - Strategy Optimization & Opponent Modeling
Built a Python research engine for Colonel Blotto, modeling opponent allocation patterns with strategy clustering and population forecasts. Implemented Numba-accelerated simulated annealing with parallel restarts to search allocations under scenario-specific payoff rules. Built walk-forward backtests and model ablations to compare predicted strategies against future opponent fields, keeping post-hoc analysis separate from forecasts.
9/2025 - 12/2025
Built a real-time ML pipeline mapping webcam input to avatar expressions using face tracking and lightweight emotion inference. Optimized streaming inference via data reduction, multithreading, and latency-first tuning, maintaining sub 5 ms end-to-end latency. Structured the system as parallel stages (capture, inference, smoothing) to avoid blocking and stabilize frame-time. Exposed a WebSocket-based real-time API enabling low-latency client synchronization and streaming.
10/2025 - 10/2025
Built a client-side Markdown compiler integrating MathJax, Mermaid, and syntax highlighting into a real-time rendering pipeline. Architected the rendering pipeline with modular abstractions to simplify future contributor onboarding and code reuse. Enhanced scalability through incremental parsing and Web Worker parallelism, sustaining sub-frame responsiveness on large documents.
9/2025 - 9/2025
Developed a Python-based automation tool that replicates server configurations between Discord communities. Utilized Discord API (discord.py v2) to manage roles, channels, and permissions with built-in rate-limit handling. Designed a config-driven, modular architecture enabling scalability and ease of extension. Implemented command-driven interaction (CLI + slash commands) and clear progress tracking for users. Emphasized API reliability, error handling, and maintainable code structure for production stability.
Chewse
8/2025 - 9/2025
Built a backend coordinating real-time shared session state between paired users using Redis. Reduced match detection from O(n) to O(1) using Redis Sets, enabling fast responses as session size scaled. Containerized services with Docker and deployed on AWS using Kubernetes and Terraform. Implemented autoscaling, health checks, and rolling updates to support reliable production deployments. Improved request throughput by optimizing concurrent request handling in Go.
7/2025 - 7/2025
Designed a concurrent Go service with isolated workers, retry/backoff, and clean context cancellation. Kept memory usage predictable under load with backpressure, bounded queues, and rate limiting. Added observability with Go pprof and runtime metrics for CPU, memory, contention, garbage collection, and worker health.
5/2025 - 5/2025
Engineered a C++17 performance engine with AVX2 SIMD, multithreading, thread-local state, and preallocated memory. Increased throughput 11x, processing 100M trials in 106 ms versus 1.19 seconds sequentially. Profiled CPU performance with Linux perf, tracking IPC, cache misses, branch misses, and cycles per trial.
1/2025 - 1/2025
Built a cross-platform testing framework in C++ to execute and validate Python algorithms with process isolation. Used Boost libraries for Inter-Process Communication and multi-threading. Implemented concurrent test execution with thread-safe result handling, reducing test suite runtimes by 60% and improving CI/CD workflows.
10/2024 - 10/2024
Worked on VisionPay, a computer vision powered self-checkout system for MarinaHacks. Designed ML infrastructure with automated preprocessing pipelines, reducing training time by 30%. Integrated backend systems with frontend for real-time recognition and receipt generation.
Mirai Research
5/2024 - 7/2024
Led development of a cryptocurrency trading platform using TypeScript and GoLang. Implemented asynchronous polling systems with circular queue-based resource management, achieving 70% faster signal detection. Built blockchain integration pipelines for reliable Solana trading operations.
Rust-Based Financial Data Aggregator
7/2024 - 7/2024
Collected and aggregated financial data through web scraping. Engineered a modular data processing pipeline using Rust, reducing processing time by 30%. Improved maintainability and extensibility through clean, modular design.
Desktop Application for Sorting Manga Work
6/2023 - 6/2023
Created a Windows application with ElectronJS to automate manga categorization using scraped metadata. Integrated results into Raindrop.io for organized management. Reduced manual sorting time by ~97%.
Chipotle Rewards Bot
11/2022 - 12/2022
Created a JS app that auto-redeemed Chipotle entree codes via SMS using VOIP. Scanned APIs and redeemed hundreds of codes, generating 4-figure value.
Ethereum Vanity Address Generator
5/2022 - 6/2022
Built a JS app that generates Ethereum wallet addresses with custom prefixes. Enabled users to obtain personalized ETH addresses by brute-force iteration.
Rias Software
8/2021 - 6/2022
Developed JS and Python tools automating retail purchases and raffle entries. Scaled concurrent task execution and optimized data scraping pipelines to handle thousands of records efficiently.