Colby Dobson
SF, CA | 415-876-8482 | colbytdobson@gmail.com | colbytdobson.com | linkedin | github
EDUCATION
University of San Francisco, San Francisco, CA
- B.S. in Computer Science August 2021 - May 2025
- Minor in Mathematics
- M.S. in Computer Science; AI Concentration August 2023 - May 2026
- President, Club Men’s Volleyball August 2023 - May 2026
WORK EXPERIENCE
Research Assistant June 2026 - Present
● Developed and evaluated a foundation model for redshift prediction across millions of astronomical spectra
● Owned and optimized experiment design, benchmarking infrastructure, data preprocessing, and scalable
multi-node, multi-GPU training pipelines on NERSC’s Perlmutter
● Built agentic workflows to automate experiment execution and evaluation, accelerating hypothesis testing
University of San Francisco, Computer Science Tutor & TA August 2024 - December 2025
● Quickly debugged and identified issues while explaining complex topics at an elementary level
Sláinte Volleyball Club, Head Coach September 2021 - July 2025
PROJECTS
Embedded Systems Arcade Game Spring 2025 - Present
● Built an ST7796S driver on the Pico SDK for RP2040, achieving 92% of theoretical max fps without visible
tearing, despite memory constrained to 2/3 a framebuffer and SPI bandwidth below one frame per refresh
● Brought up distributed Raspberry Pis with custom hardware, GPIO/UART, and real-time input arbitration
● Designed fault-tolerant timing protocols for multi-device sync with 1 controller and many players on 2 lines
Alpha-Zero Style Go Engine Spring 2026
● Built a self-play pipeline with policy/value CNNs, PUCT MCTS, and player-pool sampling from prior models
● Implemented replay tooling and candidate-vs-champion evaluation gating to compare checkpoints,
promote stronger models, and unify data generation, training, and evaluation in a continuous learning loop
Agentic Assistant Spring 2026
● Architected a multi-user agent runtime as microservices supporting task execution with tool use and
multi-step orchestration across planning, retrieval, and actions beyond single-call LLM interactions
● Built personalized RAG memory using vector databases and embedding models to persist user
preferences, retrieve long-term context, and ground agent responses and actions in user history
RISC-V Emulator Spring 2023
● Implemented a RISC-V emulator in C with instruction decoding and execution for basic instruction types,
and extended its set-associative cache simulator with multiword block fills and LRU replacement
● Built a five-stage RISC-V processor in Digital with hazard tracking, forwarding, stalls, and pipeline flushing
● Designed a dual-bank memory datapath with interleaved 64-bit banks to support misaligned byte,
word, and doubleword loads and stores across memory-word boundaries
Reverse Proxy Fall 2024
● Led a team of engineers in partnership with Consilient Labs to develop a scalable and efficient reverse
proxy as the primary layer of security for a production API with fine-grained permission control
TECHNICAL SKILLS
Programming Languages: C, C++, Python, Rust, JavaScript, Java, RISC-V Assembly
Systems: Linux, Embedded Systems, Firmware Bring-Up, GPIO/UART, IPC, Real-Time Constraints, Concurrency,
Multithreading, HW/SW Integration, Distributed Systems, Networking, Computer Architecture, Protocols
Infrastructure: Git, Docker, AWS, CI/CD, Microservices, Actions, K8s, REST, SQL, noSQL, OAuth, Prometheus/Grafana
AI/ML: PyTorch, CNNs, Transformers, Agent Workflows, Agent Infrastructure, LLM Integration, RAG, Self-Play/RL
COURSEWORK
Operating Systems, Systems Foundations, Compilers, Advanced Algorithms, Machine Learning, Bayesian Deep
Learning, Program Synthesis, Advanced Linear Algebra, Cloud Computing
