Lead machine learning engineer

• Sep 2022 - Present

Machine Learning Engineer

• Aug 2021 - Sep 2022

■ Deploy and manage infrastructure using AWS CDK (cloud development kit in python) ■ Set up a framework to build container based HTTP servers for model deployment on AWS SageMaker ■ Automate model building and testing using github actions/AWS CodeBuild, and AWS CodePipelines ■ Set up a model deployment framework using AWS SageMaker. This includes setting the instance types, scaling policies, and blue/green deployments via config files. ■ Load test model containers and scaling… Show more ■ Deploy and manage infrastructure using AWS CDK (cloud development kit in python) ■ Set up a framework to build container based HTTP servers for model deployment on AWS SageMaker ■ Automate model building and testing using github actions/AWS CodeBuild, and AWS CodePipelines ■ Set up a model deployment framework using AWS SageMaker. This includes setting the instance types, scaling policies, and blue/green deployments via config files. ■ Load test model containers and scaling policies ■ Built a data provenance workflow tool used for data engineering jobs using AWS Sagemaker processing, SageMaker Batch, AWS Fargate, AWS Batch, and AWS step functions

Senior Data Scientist/Machine learning engineer

• Sep 2020 - Aug 2021

■ Provision AWS infrastructure (Glue, Sagemaker, stepfunctions) for ML model(XGBoost) training, hyper-parameter optimisation and deployment ■ Analyse, preprocess, and extract features from data from PostgreSQL database ■ Develop serverless analytics backend on AWS with a REST API (Lambda, API gateway) ■ Develop python Dash dashboard and provision AWS infrastructure for deployment (elastic beanstalk) and authentication (Cognito) ■ Set up CI/CD pipelines with AWS CodeBuild and… Show more ■ Provision AWS infrastructure (Glue, Sagemaker, stepfunctions) for ML model(XGBoost) training, hyper-parameter optimisation and deployment ■ Analyse, preprocess, and extract features from data from PostgreSQL database ■ Develop serverless analytics backend on AWS with a REST API (Lambda, API gateway) ■ Develop python Dash dashboard and provision AWS infrastructure for deployment (elastic beanstalk) and authentication (Cognito) ■ Set up CI/CD pipelines with AWS CodeBuild and -CodePipeline for dev and prod environments Show less ■ Provision AWS infrastructure (Glue, Sagemaker, stepfunctions) for ML model(XGBoost) training, hyper-parameter optimisation and deployment ■ Analyse, preprocess, and extract features from data from PostgreSQL database ■ Develop serverless analytics backend on AWS with a REST API (Lambda, API gateway) ■ Develop python Dash dashboard and provision AWS infrastructure for deployment (elastic beanstalk) and authentication (Cognito) ■ Set up CI/CD pipelines with AWS CodeBuild and… Show more ■ Provision AWS infrastructure (Glue, Sagemaker, stepfunctions) for ML model(XGBoost) training, hyper-parameter optimisation and deployment ■ Analyse, preprocess, and extract features from data from PostgreSQL database ■ Develop serverless analytics backend on AWS with a REST API (Lambda, API gateway) ■ Develop python Dash dashboard and provision AWS infrastructure for deployment (elastic beanstalk) and authentication (Cognito) ■ Set up CI/CD pipelines with AWS CodeBuild and -CodePipeline for dev and prod environments Show less

Senior Data Scientist/Machine learning engineer

• Aug 2019 - Oct 2020

■ Train ML models to predict credit scores and affordability (XGoost, AWS Sagemaker) ■ Use genetic algorithms for feature selection and Bayesian optimisation for parameter tuning ■ Develop risk-based pricing strategies and reported on key metrics ■ Design and implement a workflow on AWS to produce risk assessments for field agents (Glue, lambda, Step functions, and SageMaker) ■ Train ML models to predict credit scores and affordability (XGoost, AWS Sagemaker) ■ Use genetic algorithms for feature selection and Bayesian optimisation for parameter tuning ■ Develop risk-based pricing strategies and reported on key metrics ■ Design and implement a workflow on AWS to produce risk assessments for field agents (Glue, lambda, Step functions, and SageMaker)

Data Scientist

• Oct 2013 - Aug 2019

■ Create detailed technical reports with results from models/algorithms (seaborn) ■ Regression analyses (linear, logistic and decision tree models) to find relationships between features or an outcome, feature significance analyses, model performance analyses (confusion matrices, kappa, F1, ROC curves) ■ Data preprocessing: unusable data is pruned, and missing data is imputed ■ Developed ML models/algorithms : motion compensated PPG-based heart rate (integrated with companies like… Show more ■ Create detailed technical reports with results from models/algorithms (seaborn) ■ Regression analyses (linear, logistic and decision tree models) to find relationships between features or an outcome, feature significance analyses, model performance analyses (confusion matrices, kappa, F1, ROC curves) ■ Data preprocessing: unusable data is pruned, and missing data is imputed ■ Developed ML models/algorithms : motion compensated PPG-based heart rate (integrated with companies like Garmin, TomTom, and Montblanc), device-on-skin detection, exercise-activity classification, sleep apneoa classification, and drowsiness detection ■ Python machine learning packages used : scikit learn, statsmodels, imblearn, patsy, pyGAM and keras (Tensorflow) ■ ML models used : linear- and logistic regression, SVM, decision trees, ensemble models (gradient boosting, random forests), feature selection (Pearson’s correlation, Spearman’s correlation, mutual information, PCA, ICA, recursive feature elimination, LASSO based, tree based, etc.), deep learning (CNN, LSTM), and clustering/Gaussian mixture models ■ Frequency- and time domain analyses on time series data for feature extraction ■ Time series analysis techniques used : FFT, DWT, Lombscargle, sparse signal reconstruction, filter design (FIR, IIR), adaptive filters (notch, comb, Kalman, LMS, RLS, QRLS), PCA, ICA, emprical mode decomposition, detrended fluctuation analysis, hurst exponent, correlation dimension, and sample entropy Show less ■ Create detailed technical reports with results from models/algorithms (seaborn) ■ Regression analyses (linear, logistic and decision tree models) to find relationships between features or an outcome, feature significance analyses, model performance analyses (confusion matrices, kappa, F1, ROC curves) ■ Data preprocessing: unusable data is pruned, and missing data is imputed ■ Developed ML models/algorithms : motion compensated PPG-based heart rate (integrated with companies like… Show more ■ Create detailed technical reports with results from models/algorithms (seaborn) ■ Regression analyses (linear, logistic and decision tree models) to find relationships between features or an outcome, feature significance analyses, model performance analyses (confusion matrices, kappa, F1, ROC curves) ■ Data preprocessing: unusable data is pruned, and missing data is imputed ■ Developed ML models/algorithms : motion compensated PPG-based heart rate (integrated with companies like Garmin, TomTom, and Montblanc), device-on-skin detection, exercise-activity classification, sleep apneoa classification, and drowsiness detection ■ Python machine learning packages used : scikit learn, statsmodels, imblearn, patsy, pyGAM and keras (Tensorflow) ■ ML models used : linear- and logistic regression, SVM, decision trees, ensemble models (gradient boosting, random forests), feature selection (Pearson’s correlation, Spearman’s correlation, mutual information, PCA, ICA, recursive feature elimination, LASSO based, tree based, etc.), deep learning (CNN, LSTM), and clustering/Gaussian mixture models ■ Frequency- and time domain analyses on time series data for feature extraction ■ Time series analysis techniques used : FFT, DWT, Lombscargle, sparse signal reconstruction, filter design (FIR, IIR), adaptive filters (notch, comb, Kalman, LMS, RLS, QRLS), PCA, ICA, emprical mode decomposition, detrended fluctuation analysis, hurst exponent, correlation dimension, and sample entropy Show less

External Examiner

• Jan 2018 - Apr 2019