Profile Summary
Currently working as an individual contributor in the internal filtering platform team in Google. Holds an M.S. Degree from UC San Diego specializing in Machine Learning and Data Science.
My professional interests include software development, distributed systems and machine learning.
Professional History
Architect and scale the core Filtering Platform to enable internal clients, such as Gemini, to enforce legal and regulatory compliance.
Collaborate cross-functionally with engineering, research, and legal stakeholders to design and deploy robust, low-latency data policy guardrails.
Engineered distributed indexing pipelines processing petabytes of web data, improving the system to purge low-value and duplicate pages to measurably elevate search result quality.
Optimized data pipelines to accelerate search ranking signal refresh cycles from weeks to days, drastically improving data freshness.
Built SageMaker Processing APIs and orchestration layers for distributed ML tasks.
Engineered API integrations connecting SageMaker processing services with AWS Athena, AWS Redshift, and Snowflake
Developed SageMaker backend services for large-scale training and high-throughput inference.
Delivered instruction, managed code evaluations, and designed exercise sessions covering Statistical Learning math paradigms (ECE-271 A) and Pattern Recognition algorithm implementations (ECE-175 A).
Built spatial scale inference modules connecting a Mask R-CNN foundation directly into a parallel Siamese deep learning layout to deduce target structural variables. Managed crowdsourced parsing arrays using Mechanical Turk infrastructure to acquire bespoke validation datasets.
Constructed data evaluation engines tracking 9-axis spatial IMU systems to map biomechanical stroke profile movements. Built isolated micro-state analysis layers identifying real-time physical stress markers via multi-channel EDA and HRV signals.
Academic Research & Publications
- Compressive sensing for power efficient data aggregation in a wireless sensor network (US Patent)
- Physiological Signal Based Work Stress Detection Using Unobtrusive Sensors
- Differential effects of physical and psychological stressors on electrodermal activity
- Design of a 1V subthreshold comparator for bio-medical over sampled ADCs
Projects
Captioning & Retrieval
Coupled ResNet-152 visual layers with recurrent LSTM cells to execute complex sequential context parsing and automated contextual item matching.
Mathematical Parser
Constructed structural character localization networks paired with Random Forest classifiers to segment and parse handwritten equations.
IMDB Metrics Engine
Evaluated extensive graph data configurations to compute demographic analytics and build fast collaborative item-filtering recommender clusters.
Contextual Classifier
Implemented TF-IDF transformation matrices alongside Latent Dirichlet Allocation models to track text anomaly vectors.
Matrix Recommendation
Applied matrix factorization and hidden factor models to predict user preference profiles within dense commercial networks.
Universal Portfolio Engine
Coded recursive multi-asset portfolio balancing loops derived from Cover's Universal Portfolio formulations to evaluate historical data.