joon lee | about

physical intelligence

experiences

I love using physical machine learning, intelligent robotics, and spatial computing to engineer delightful products.

machine learning + product design @ samsung

I prototyped Gemini’s circle to search interactions for Android XR by training YOLO object detection models on eye, gesture, and speech inputs to rank 16K+ objects as search results in 12 languages for 80K+ users.

machine learning + product design @ meta

I designed Meta Ray-Ban's Neural Band gesture interactions by training EMG signal classifiers on wrist contractions to spatially map hand gestures to live navigation commands in 16+ languages for 1M+ users.

machine learning + product design @ apple

I redesigned visionOS writing tools and smart reply interaction model by training LLM agents on eye, gesture, and optic ID inputs to deliver writing assistance in 12+ languages for speech-impaired users.

machine learning + robotics @ google

I prototyped Gemini's vision-language-action model using PaLM-2 LLMs, embodied 3D spatial reasoning, and physical affordance probabilities to translate camera and language inputs into robotic motor actions.

Overlapping photo cards displaying a hand holding a green smartphone with a crop analytics dashboard above a pile of fresh vegetables including tomatoes and mushrooms against a blue sky, a smartphone showing an avocado plant disease detection screen beside a whole avocado on a stone pedestal against a cream background, and a smiling farmer in denim overalls and a bucket hat standing in a green agricultural field, arranged on a light gray background with a dark teal organic blob shape and a mint green star-shaped accent.

agsight

I designed a geospatial AI app to deliver predictive agricultural insights to 50+ vineyards and orchards in 4 countries, helping 4 Napa Valley winegrowers forecast smoke taint during the 2025 California wildfires.

small questions, big answers

research

I love pursuing questions at the intersection of environmental and computational intelligence.

machine learning + physics @ nasa

I improved the accuracy of probabilistic global weather forecasts for tropical cyclones, atmospheric rivers, and heat waves in atmospheric physics simulations by training graph neural networks on satellite images.

Overlapping photo cards showing a red PCA biplot of North American fungal isolates clustered by principal components, a photograph of the NASA Ames Research Center water tower, and a wildfire burn confidence probability curve comparing logistic regression and CNN model outputs, with a standard deviation formula on a light gray background with a pink blob shape.

machine learning + physics @ nasa

I engineered a distributed wildfire early-warning system to produce probabilistic fire spread maps by using Monte Carlo physics simulations to quantify wildfire sensitivity to vegetation, topography, and wind.

Overlapping photo cards showing a GAM regression curve modeling phosphorus cycling over time, a damped oscillation waveform with error function equation, and a photograph of Gates Hall at Cornell Tech, on a light gray background with a Cornell University seal and abstract red accent shapes.

machine learning + spatial computing @ cornell

I designed spatial AR user interfaces for autonomous healthcare robots to project medical data onto AI clinician headsets using computer vision, DBSCAN clustering, and 3D kinematics, sponsored by Google.

Overlapping photo cards showing a bar chart and violin plot comparing microplastic abundance distributions across ocean regions, a global microplastic concentration map with color-coded density gradients, and a stochastic neural network convergence curve alongside a hyperspectral heatmap, on a light gray background with a Cornell University seal.

electrical + computer engineering @ cornell

I prototyped an autonomous AI robot to diagnose grapevine leafroll diseases and phenotype powdery mildews for 12+ New York vineyards using hyperspectral imaging and thermal sensors, funded by NASA.

electrical + computer engineering @ yale

I engineered an AI robot by researching how piezoelectric transducers convert electric potential energy into 3D kinematic motion guided by state estimators, hierarchical motion planning, and reinforcement learning.