Ben Huang

Skills

Experience

1. TODO: Start – end, e.g. Jun 2025 – Aug 2025

   TODO: Job title, e.g. Software Engineering Intern

   TODO: Company or lab name, e.g. Acme Robotics

   • TODO: What you built or owned, and the technology you used.
   • TODO: A measurable result — latency, throughput, cost, users, or scope.
   • TODO: Optional third point — collaboration, ownership, or something shipped.
2. TODO: Start – end, e.g. Jan 2025 – May 2025

   TODO: Second role — e.g. Undergraduate Researcher, or delete this entry

   TODO: Organization, e.g. a UC Berkeley lab or a student org

   • TODO: What the work was and your specific contribution to it.
   • TODO: The outcome — a paper, a tool others use, a demo, a result.

Projects

• 
  Worldcraft(opens in a new tab)

  Turns ordinary photographs into playable Minecraft worlds using computer vision and neural radiance fields.

  Worldcraft runs a four-stage pipeline: semantic segmentation with Mask2Former labels the scene, Nerfstudio trains a NeRF to recover 3D structure, the resulting point cloud is voxelized, and each voxel is mapped to a Minecraft block. The interesting part is bridging continuous neural geometry with a hard voxel grid — segmentation labels carry through the reconstruction so the final world keeps materials that match the original scene.

  • Python
  • PyTorch
  • NeRF
  • FastAPI
• 
  GPGraph: GPU Graph Processing(opens in a new tab)

  A CUDA library that moves classic graph algorithms onto the GPU for high-throughput processing.

  GPGraph is written in CUDA C/C++ and implements graph traversal and analysis routines as data-parallel kernels rather than pointer-chasing CPU loops. Work on this kind of library lives and dies on memory behaviour — coalesced access over the adjacency structure, keeping thread divergence low on irregular neighbour lists, and balancing work across warps when vertex degrees are skewed.

  • Cuda
  • C
  • C++
• 
  DECODE(opens in a new tab) 2025

  A bioinformatics tool built at the 2025 QBI Hackathon that applies genomic foundation models to biological sequence data.

  DECODE is written in Python on top of PyTorch and uses Evo2, a genome-scale sequence foundation model, to score and interpret biological sequences. Building on a foundation model of that size at hackathon pace meant the work sat mostly in the data path — getting sequences into the right representation, running inference within the available GPU memory, and turning raw model outputs into something a biologist could read.

  • Python
  • Torch
  • Evo2
• 
  Hashrate Mining Optimizer(opens in a new tab)

  A reinforcement learning system that predicts and allocates mining hashrate from operational data.

  Built in Python with PyTorch for the MARA hackathon, the project frames hashrate allocation as a sequential decision problem: an agent observes operating conditions and learns a policy for how much hashrate to commit, rather than reacting with fixed thresholds. Mining is a natural fit for RL because the trade-off between energy cost and yield shifts continuously, so a learned policy can adapt where static heuristics cannot.

  • Python
  • Torch
• 
  WattVision(opens in a new tab) 2026

  Scan your home appliances with on-device AI to track energy usage and catch ghost power drain.

  Built at SF Hacks 2026, WattVision pairs a TypeScript React Native and Expo client with a Python FastAPI backend, running appliance recognition on-device so a scan works without shipping camera frames to a server. One codebase ships to web, Android, and iOS, and identified appliances are turned into an energy picture that surfaces the always-on loads people never think to look for.

  • TypeScript
  • React Native
  • Python
  • FastAPI

Contact

I'm always happy to talk about interesting engineering problems, internships, or new-grad roles. The fastest way to reach me is email.

• benhua476@berkeley.edu
• github.com/benhuang3