B.S. Electrical & Computer Engineering @ Carnegie Mellon University
Minor in Machine Learning | Class of 2026
🎓 I'm a systems and ML engineer passionate about building efficient, scalable, and intelligent infrastructure — from distributed protocols and real-time kernels to deep learning frameworks and generative models.
💡 Currently, I’m a Research Assistant at CMU, where I:
- 🧩 Develop LoRA-based fine-tuning pipelines for compact open-weight LLMs (LLaMA 3 8B)
- 🧠 Design dynamic model routing systems using contextual bandit algorithms
- 📊 Benchmark fine-tuning efficiency across adapter architectures
💼 Previously interned at:
- Capital One (Summer 2025) – built scalable performance-testing infrastructure for 15+ microservices (10K+ TPS, 5 ms latency)
- Ansys (Spring 2025) – optimized CMake builds and CI pipelines for mechanical simulation software
- Efficio (Summer 2024) – designed cloud-cost optimization tools reducing deployment costs by 23%
🧠 Mini Deep Learning Framework (Python + C++)
A PyTorch-like mini framework built from scratch featuring:
- Reverse-mode autograd and dynamic computation graphs
- Custom CUDA kernels for tensor ops → 20× faster than CPU
- Modular API for model definition and training
🧩 Technologies: Python, C++, CUDA, NumPy, PyTorch Autograd internals
⛏️ Distributed Bitcoin Miner (Go)
- Designed the Live Sequence Protocol (LSP) for reliable UDP communication
- Implemented concurrency with goroutines + channels for scalable task distribution
- Added fault-tolerant load balancing and miner reassignment logic
🧩 Technologies: Go, Networking, UDP, Concurrency, Distributed Systems
⏱️ Real-Time Operating System (C + ARM Assembly)
- Built a real-time kernel supporting context switching, task scheduling, and mutexes
- Implemented rate-monotonic scheduling (RMS) with schedulability checks
- Designed thread control blocks (TCBs) for multi-threading management
🧩 Technologies: C, ARM Cortex-M, Embedded Systems, RTOS Design
🌀 Denoising Diffusion Probabilistic Model (PyTorch)
- Implemented forward + reverse diffusion processes using cosine variance scheduling
- Trained U-Net-based DDPM on Animal Faces-HQ for image generation
- Evaluated using Fréchet Inception Distance (FID) and visualized training convergence
🧩 Technologies: PyTorch, Diffusion Models, Generative AI, Deep Learning
- 🔬 Research on efficient LoRA fine-tuning and adapter composition at CMU
- 🧠 Expanding my custom DL framework with quantization + distributed training
- ☁️ Experimenting with cloud-native LLM deployment pipelines using Docker + AWS
If you’re interested in ML systems, low-level optimization, or AI infrastructure,
feel free to reach out — I love collaborating and discussing new ideas!