# Edward Chen — Resume San Francisco, California · 925-725-5285 · echen1246@gmail.com GitHub: https://github.com/0x3ddie · Portfolio: https://eddiechen.xyz/ Canonical PDF: https://eddiechen.xyz/Edward-Chen-Resume.pdf --- ## Education **Arizona State University** — August 2023 to present BS Data Science, BA Supply Chain Management, Minor in Economics. GPA: 3.7. Coursework: Data Structures, Operating Systems, Computer Vision, Networking, Machine Learning, Linear Algebra. Activities: Scholars of Finance, Association for Computing Machinery. --- ## Projects ### KV-Cache Compression for Transformer Inference — `local-turboquant` *PyTorch, Modal, Transformers, CUDA, Triton* - Reproduced an unpublished ICLR 2026 paper from scratch and shipped it as a working library before the authors released code. One-line API, Python wrapper, and a CLI exposing per-generation savings, throughput, and freed VRAM. Benchmarked on H100s via Modal. - Wrote a fused Triton attention kernel that runs directly on bit-packed 3-bit KV indices without decompressing to FP16, achieving ~74% KV cache memory reduction at near-lossless quality. Repo: https://github.com/0x3ddie/local-turboquant ### Modified Transformer Architecture — `smarternano` *PyTorch, CUDA, OpenRouter, Transformers* - Re-engineered the nanochat architecture with custom layer depth and weight initialization, re-trained on Nvidia's Nemotron Nano dataset; the project was publicly recognized by Andrej Karpathy. - Built a synthetic data generation pipeline via OpenRouter to align the model's personality, achieving coherence comparable to 560M parameter models. Demo: https://space3--nanochat-serve-chat.modal.run/ ### Wasm Runtime and Compiler — `silicon-JIT` *Rust, ARM64 Assembly* - Engineered a WebAssembly runtime in Rust that executes C/Rust programs on Apple Silicon, with a JIT compiler translating bytecode to native ARM64 machine code. - Managed executable memory via manual `mmap`/`mprotect` syscalls for direct code generation. ### AI-Native App — `murmur` *ONNX, Flutter, AI Engineering* - Built and released a mobile app that converts any PDF into human-quality audio entirely on-device, achieving 3-second load-to-playback with zero server dependencies. Deployed Kokoro-82M transformer TTS on Android via ONNX Runtime, forking the runtime to isolate inference on a background thread for performance. - Built an NLP preprocessing pipeline by cross-compiling espeak-ng and writing Dart FFI bindings for IPA phonemization, implementing token-aware batch splitting against the model's tokenizer vocab, and caching per-sentence phonemes in SQLite to eliminate redundant compute at inference time. App: https://play.google.com/apps/testing/com.nonchalant.murmur --- ## Experience ### mymelo.org — Co-Founder, ML Engineer *April 2025 to present* - Trained and deployed an EfficientNet-B0 CNN on 30K dermoscopy samples achieving 95% diagnostic accuracy, serving inference via AWS Lambda with an S3-backed data pipeline. - Built a cross-platform mobile app in Flutter with OAuth integration, conducting 50+ user interviews to iterate on UX for clinical deployment. - Engineered a scalable data pipeline using AWS S3 and Lambda to support a high-volume image processing workflow for model inference and data archival. ### M.Y Intellectual Property — Software Engineer Intern *May 2024 to August 2024* - Engineered a unified patent search API aggregating USPTO, CNIPA, and EPO databases into a single query interface, reducing cross-jurisdictional research from 3 workflows to 1 and cutting attorney lookup time by 60%. - Built a citation network parser to extract patent family relationships across 50K+ international filings, automating competitor portfolio analysis for 15+ client engagements. --- ## Skills - **Languages:** C++, Python, Java, JavaScript, Rust, SQL. - **Technologies:** AWS, Docker, Flutter, React, Next.js, Django, Flask, Supabase, Git, MongoDB, Vue.js, SQLite, Modal. - **AI/ML:** PyTorch, FastAPI, ONNX Runtime, OpenCV, CUDA, vLLM, LangChain, FAISS, ANOVA, Keras, Transformers. - **Open-source contributions:** Meta PyTorch. --- ## Links - Website: https://eddiechen.xyz/ - Projects: https://eddiechen.xyz/engineering - Writing: https://eddiechen.xyz/writing - GitHub: https://github.com/0x3ddie - X / Twitter: https://twitter.com/ayocheddie