• Design (SolidWorks, Fusion 360) and fabricate planting mechanism that is attached to ground robot to dig holes, deposit seedlings, and securely close holes autonomously

• Prototype and reiterate designs with 3D printing, laser cutting, and machining

• Performed GIS study to identify pasture land in US to quantify reforestation opportunities for ground robot

• Program microcontroller to send commands to various motors that are connected to planting mechanism to perform planting procedure in correct sequence and make use of current feedback from motor that drills holes for safety stop

• Integrate optical depth sensor to help planting subsystem detect ground and determine where to plant

• Implemented MPC for ground robot to follow planned route of planting seedlings- used differential drive kinematics equations for dynamics constraint and integrated obstacle avoidance in inequality constraints

• Install limit switches to provide feedback to planting mechanism for proper system resets

Experimentation, Inference, and Product Analytics

• Design and analyze experiments to improve Consumer facing products - e.g. web to app upsell, personalized onboarding process, restaurant ranking algorithms, exclusive offer carousels

• Perform exploratory analysis to identify new opportunities utilizing support from both data and psychological concepts (e.g. cognitive load, social proofing, exploration and exploitation)

• Utilize causal inference techniques when experimentation is not possible or too costly

• Conduct deep dives on experiment results to guide iterative feature development

• Work closely with product managers and engineers to help steer team's direction

• Developed music recommender system for JOOX, Southeast Asia’s popular on-demand music streaming platform

• Collaborative filtering: implemented and evaluated models that used matrix factorization (ALS) and k-nearest neighbors

• Performed music information retrieval to extract audio features such as MFCC, chromagram, tempo, and zero crossing rate and used the features to automatically tag songs with attributes (e.g. genre, mood, valence) using standard classifiers (random forest, SVM) and deep learning (mainly CNN on spectrograms)

• Built song similarity model (combination of KNN and association rule learning) for Thai music, based on audio tags, lyrics content, and user feedback, behavior, and profile

• Worked closely with engineers to deploy recommender system into production and to implement event tracking to ensure data is properly collected to perform user behavior analytics

• Designed and managed A/B testing experiments to evaluate different models and UX/UI using Experiments with Google

• Created, deployed, and maintained JOOX Analytics dashboard, which contained real-time song, music video, advertisement, and campaign performance, to enable business and operation teams to make data-driven decisions

• Analyzed user behavior data to find key insights that help understand customers better. For example: analysis on new hip-hop trend that doesn't appeal to certain groups, marketing strategy based on distinct geographical music preferences