I worked as a Data Science Engineer in Viken Detection, Burlington
I worked as a Data Science Engineer in Viken Detection, Burlington
Embedded Files
The flowchart of my work to classify narcotics drugs using Machine Learning Algorithms and deployed to production. Made with LucidChart.
The flowchart of my work to classify narcotics drugs using Machine Learning Algorithms and deployed to production. Made with LucidChart.
Handheld Chemical Analyzer for Narcotics drug classification.
Handheld Chemical Analyzer for Narcotics drug classification.
Foxhound has been an important tool used by law enforcement to help protect the public from the opioid crisis.
Foxhound has been an important tool used by law enforcement to help protect the public from the opioid crisis.
Last year, 107,375 people in the U.S. died of drug overdoses, with 67% of those deaths due to synthetic opioids.
Last year, 107,375 people in the U.S. died of drug overdoses, with 67% of those deaths due to synthetic opioids.
Drug Classification Algorithm
Drug Classification Algorithm
A supervised Machine Learning algorithm was crafted for a chemical analyzer to effectively categorize drugs utilizing ensemble and boosting techniques, resulting in 96% overall accuracy in narcotics multi-class classification.
A supervised Machine Learning algorithm was crafted for a chemical analyzer to effectively categorize drugs utilizing ensemble and boosting techniques, resulting in 96% overall accuracy in narcotics multi-class classification.
The algorithm was carefully designed to maintain a false alarm rate of under 3%, showcasing the resilience and dependability of the machine learning methodologies utilized.
The algorithm was carefully designed to maintain a false alarm rate of under 3%, showcasing the resilience and dependability of the machine learning methodologies utilized.
Peak detection and feature engineering methodologies were applied to scrutinize tabular data containing 80,000 data points from a repository of 15 drugs, facilitating the identification of distinct characteristics for each drug.
Peak detection and feature engineering methodologies were applied to scrutinize tabular data containing 80,000 data points from a repository of 15 drugs, facilitating the identification of distinct characteristics for each drug.
Classical ML - Ensemble and Boosting Techniques
Deployment and Optimization
Deployment and Optimization
Contributed to deploying this classification model into production on an edge device and developed user-friendly features for the application using Android Studio.
Contributed to deploying this classification model into production on an edge device and developed user-friendly features for the application using Android Studio.
Optimized the application to achieve an overall run-time of 11 seconds, with drug detection taking only 3 seconds. This optimization enhances the usability and practicality of the algorithm in diverse field conditions, ensuring swift and accurate prediction.
Optimized the application to achieve an overall run-time of 11 seconds, with drug detection taking only 3 seconds. This optimization enhances the usability and practicality of the algorithm in diverse field conditions, ensuring swift and accurate prediction.
Collaborated with the team members to implement, test, and debug various model versions using Git to ensure the target accuracy was achieved.
Collaborated with the team members to implement, test, and debug various model versions using Git to ensure the target accuracy was achieved.
Research and Development (R&D)
Research and Development (R&D)
Conducted extensive research to explore the algorithm's adaptability to variations in pressure and humidity as these factors affect the prediction.
Conducted extensive research to explore the algorithm's adaptability to variations in pressure and humidity as these factors affect the prediction.
With new findings, I helped calibrate the device to enhance sensitivity to low-signature drugs like Heroin and THC (Marijuana), resulting in a significant improvement in the overall accuracy.
With new findings, I helped calibrate the device to enhance sensitivity to low-signature drugs like Heroin and THC (Marijuana), resulting in a significant improvement in the overall accuracy.
Using SHAP, I presented clear statistical analyses of experiments to stakeholders for interpretation and feedback. (SHAP is a game theory-based approach to explain the output of machine learning models.)
Using SHAP, I presented clear statistical analyses of experiments to stakeholders for interpretation and feedback. (SHAP is a game theory-based approach to explain the output of machine learning models.)
This experience contributed to the ongoing efforts in the fight against drug trafficking using machine learning algorithms. This instilled in me a strong sense of purpose and reinforced my belief in the power of data to make a difference.
This experience contributed to the ongoing efforts in the fight against drug trafficking using machine learning algorithms. This instilled in me a strong sense of purpose and reinforced my belief in the power of data to make a difference.
Page updated
Google Sites
Report abuse