Hey there, I'm

Calvin Zhou

I build systems that perceive, reason, and act, from autonomous drones to agentic search.

NowEng @ Bodhium Labs

ResearchPenn PAL Group

Selected projects

A2D2

Autonomous Wildfire Detection

A drone network that finds wildfires before they spread. Deep computer vision, coverage-path planning, and deep RL in one pipeline.

Validated by CalFire, USFS & NASA

Deep RLComputer VisionAlgorithms

Paper ↗ Code ↗

FERRO

Real-time Policy Serving for Robots

A two-rate C++ runtime that serves an RL policy to a simulated 7-DOF arm. A 1 kHz control loop stays steady even when inference stalls.

35.5µs control p99 under 3x policy stalls

C++RustReal-time

Code ↗

Ferro: 7-DOF arm pushing a puck onto a goal

AIBEE

Block-coding AI Classroom

A platform that teaches AI by letting students build it. Live Python, a code to blocks compiler, and an AI tutor.

Piloted with 20+ students

Full-stackMCPKernels

Talk ↗ Paper ↗ Code ↗

CRIT

Critique-driven Reward Models

A reward model for UI design. LLM judges score rendered screenshots across critique loops, yielding a failure taxonomy for fine-tuning.

RL EnvsLLM EvalsFine-tuning

Code ↗

Research

As a research assistant in Penn's Perception, Action & Learning group, I've worked on world-model architectures and RL policy evals for dexterous manipulation. Currently, I'm designing guidance for flow-matching policies and building training and eval pipelines for VLAs.

Nov 2025 – Present
Philadelphia, PA

Flow matchingGuidanceVLAsWorld modelsRL

As a researcher at Bodhium Labs, I work on web-scale data, AI search, and knowledge extraction.

Web-as-CorpusNLPLLM Evals

Publications & Talks Google Scholar ↗

You Can Tell Who's Asking: What the Web's Questions Are Made Of, and Where They Come From WAC-13 @ EMNLP 2026 · Oral ↗ A2D2: AI-Driven Autonomous Drone-Based Detection Network for Wildfires ARIS 2024 · Taipei ↗ AI-Driven Block-coding Educational Environment for Teaching, Learning, and Prototyping AI Talk · MIT AI + Edu Summit 2024 ↗

Experience

Dec 2025 — Present
Menlo Park, CA

Bodhium Labs

Software and Research Engineer

Applied AI research lab startup led by Wharton prof and former Deepmind eng.
A lot of building, learning, researching, shipping, and working with clients.

Summer 2025
Lexington, MA

MIT Lincoln Laboratory · Group 105

Research Intern

Worked on signal processing, radar, and drones. Met some insanely smart and kind people.

2021 - 2025

W3B Education

Founder

Founded and ran a youth entreprenuership org. Reached 2,000+ students across 30+ countries.
Talked to some really cool people and met a lot of my mentors.

Sep 2025 - Apr 2026
Philadelphia, PA

Wharton Impact Venture Associates

Analyst

Talked to many early-stage climate tech startup founders and led a deal.

Writing

Time to start writing more. I guess this is the beginning... After a while, I realized paying for a domain should mean that this website should be more than just my resume in a different font :) ↗

About

I'm an engineer and researcher at the intersection of AI and robotics. I like building systems that touch the real world, and I think best at the boundary between research and shipping. Penn M&T (AI, CS, and Finance)

Languages: Python · C++ · Rust · TypeScript · JavaScript · Java · OCaml · HCL
ML / AI: PyTorch · JAX · TensorFlow · ONNX · MuJoCo · Hugging Face · OpenCV
Systems: FastAPI · React · Next.js · Docker · PostgreSQL · Terraform
Cloud: GCP · AWS