Kamohelo Thuntsi

Experience

• 

  Walmart

  Apr 2022 - Aug 2022

  Data Scientist

  Responsibilities:⬩ Designed and developed a sales forecasting system for Walmart⬩ Maximized accuracy of demand projections considering events and holidays' impact on retail sales patterns⬩ Used variables such as Consumer Price Index (CPI), Unemployment Index, and seasonal promotional markdown events⬩ Identified stores with highest sales, sales standard deviation, and significant quarterly growth⬩ Analyzed holidays with higher sales compared to non holiday seasons and monthly/semester sales trends⬩ Developed a statistical model using Linear Regression tailored for Store 1, incorporating variables like date, CPI, unemployment, and gas prices⬩ Maximized model accuracy by representing dates as days⬩ Implemented system resulted in significant revenue gains, improved inventory management, reduced stockouts, and enhanced customer satisfaction for Walmart⬩ Leveraged insights on holidays and economic conditions to optimize promotional markdowns and adjust sales strategies effectively⬩ Contributed to improved sales performance, streamlined operations, and maximized competitiveness in the retail sector through decision making processes based on dataTech stack: Python, pandas, scikit-learn, Jupyter Notebook Show less

• 

  Uber

  Apr 2022 - Jul 2022

  Artificial Intelligence Engineer

  Responsibilities, What I did:⬩ Designed and developed a fare prediction algorithm for Uber⬩ Gathered historical ride data including distances, durations, pickup/drop-off locations, timestamps, and fare amounts⬩ Engineered features and extracted insights to improve prediction model accuracy⬩ Calculated estimated trip durations, incorporated traffic data, and encoded categorical variables⬩ Evaluated and selected machine learning algorithms, choosing Gradient Boosting Regressor for superior performance and flexibility⬩ Trained the model on prepared dataset, fine-tuning parameters and optimizing performance using cross-validation⬩ Conducted hyperparameter tuning using grid search and randomized search methods to optimize learning rate, tree depth, and number of estimators⬩ Evaluated model performance using metrics such as mean absolute error (MAE), root mean squared error (RMSE), and R-squared (R2) score⬩ Contributed to Uber's objective of delivering transparent and accurate fare estimates, improving overall user experience.Tech stack: Python, pandas, scikit-learn, XGBoost, Jupyter Notebook Show less

• 

  Twitter

  May 2022 - Aug 2022

  Machine Learning Engineer (Natural Language Processing - NLP)

  Responsibilities:⬩ Designed and developed a model to identify and remove hate speech on Twitter / X⬩ Conducted data cleanup including normalization, removal of user handles, URLs, stop words, and redundant terms⬩ Used TweetTokenizer from NLTK for tokenization and specific cleanup tasks⬩ Employed TF-IDF values as features for predictive modeling⬩ Initially used Logistic Regression and addressed class imbalance to ensure equal focus on hate and non hate speech⬩ Applied regularization and hyperparameter tuning using techniques like GridSearch and StratifiedKFold⬩ Prioritized recall as the scoring metric due to class imbalance in hate speech detection⬩ Evaluated model performance using metrics such as accuracy, recall, and f1 score⬩ Aimed to create a robust model for effectively identifying and mitigating hate speech on Twitter⬩ Contributed to maintaining a safe and positive online environment for Twitter / X users⬩ Enabled efficient content moderation and upheld community guidelines for a more inclusive online communityTech stack: Python, pandas, scikit-learn, TensorFlow/Keras, Jupyter Notebook Show less

• 

  LendingClub

  Jun 2022 - Sept 2022

  Machine Learning Engineer

  Responsibilities:⬩ Developed a deep learning model to predict loan defaults using past data from Lending Club⬩ Conducted data preparation by converting categorical values into discrete numerical values⬩ Performed exploratory data analysis (EDA) to uncover relationships and identify important features⬩ Implemented feature engineering strategies to enhance dataset quality⬩ Assessed feature correlations and removed redundant features to improve model performance⬩ Built a deep learning model with Keras and TensorFlow backend to predict loan default probabilities⬩ Provided valuable insights into loan default risk for Lending Club⬩ Facilitated better informed financial decision making process through predictive analyticsTech stack: Python, pandas, scikit-learn, TensorFlow/Keras, Jupyter Notebook Show less

• 

  Mercedes-Benz AG

  Aug 2022 - Oct 2022

  Machine Learning Engineer

  Responsibilities:⬩ Designed and developed a Test Bench Optimization System for Mercedes-Benz⬩ Conducted variance check to remove columns with zero variance, improving data relevance⬩ Ensured data integrity by assessing quality through null value and unique entry checks⬩ Implemented label encoding to transform categorical data into numerical format for machine learning⬩ Employed dimensionality reduction techniques to simplify data and maximize model efficiency⬩ Made use of the XGBoost algorithm for accurate forecasting of test data values⬩ Reduced test bench time, leading to faster production cycles and decreased carbon dioxide emissions⬩ Maintained high standards of quality and safety while improving operational efficiency⬩ Supported Mercedes Benz's commitment to advancement, effectiveness, and environmental responsibilityTech stack: Python, pandas, scikit-learn, XGBoost, Jupyter Notebook Show less

• 

  Amazon

  Sept 2022 - Dec 2022

  Data Scientist

  Responsibilities:⬩ Designed and developed a sentiment analysis system for Amazon's customer reviews⬩ Conducted exploratory data analysis (EDA) to understand the dataset and address class imbalance⬩ Transformed reviews into Tf-Idf ratings and used Multinomial Naive Bayes for sentiment prediction⬩ Evaluated models using metrics like F1-Score to assess performance⬩ Explored model augmentation and selection methods including multi-class SVMs, neural networks, and ensemble approaches⬩ Implemented deep learning models such as LSTM and GRU to compare with traditional machine learning algorithms⬩ Used topic modeling techniques like Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF) to identify recurring topics in reviews⬩ Provided Amazon with insights to maximize customer satisfaction through quick issue resolution⬩ Assisted in improving product quality by identifying recurring review topics⬩ Optimized marketing strategies by targeting specific customer needs identified through sentiment analysis⬩ Contributed to Amazon's competitive edge in the e-commerce market through informed decision making and business growth.Tech stack: Python, pandas, scikit-learn, TensorFlow/Keras, NLTK, Gensim, Jupyter Notebook Show less

• 

  Comcast

  Nov 2022 - Jan 2023

  Data Analyst

  I contributed to the development of a Customer Service Optimization System for Comcast, focusing on enhancing their complaint resolution workflows. I designed and implemented a machine learning model that analyzed and structured critical complaint data, including complaint IDs, descriptions, timestamps, communication channels, customer locations, statuses, and proxy complaints. This allowed for deeper insights into customer pain points and service gaps, providing Comcast with actionable intelligence to optimize their service operations.I developed a classification model to categorize complaints into “Open” and “Closed” states, streamlining the prioritization process and resource allocation for resolution efforts. This feature significantly improved operational efficiency by ensuring that unresolved complaints were promptly addressed. The model also employed advanced data analysis techniques to generate frequency tables, categorizing complaints by issue type (e.g., internet outages, network problems) to identify the most prevalent and impactful issues.To improve customer service response times and resolution strategies, I incorporated time-series analysis and visualization tools. These tracked complaint volumes on a daily and monthly basis, enabling Comcast to pinpoint high-complaint periods and adjust their operational readiness. And, I implemented state-wise analysis of complaint statuses, identifying regions with higher complaint volumes and supporting targeted interventions for customer satisfaction.Lastly, I designed the system to calculate resolution success rates by differentiating between issues resolved via online channels and those handled by customer care, giving Comcast a clear overview of their service performance. The insights generated by the model enabled them to evaluate and refine their customer service channels for maximum effectiveness.Tech stack: Python, Pandas, NumPy, Scikit-learn, Matplotlib, Seaborn, SQL Show less

• 

  Inter-American Development Bank

  Feb 2023 - Apr 2023

  Machine Learning Engineer

  Responsibilities:⬩ Designed and developed an income qualification system for a social assistance program in Latin America⬩ Implemented the Proxy Means Test (PMT) method using algorithms to determine income qualification based on observable household factors⬩ Used data containting Costa Rican household characteristics to maximize accuracy of the PMT model⬩ Identified the output variable and conducted comprehensive data assessment to ensure data accuracy⬩ Addressed biases and inconsistencies in the dataset to maintain data integrity⬩ Managed complexities such as varying poverty levels among household members and households without a designated family head⬩ Conducted thorough analysis to detect and handle null values, improving data reliability⬩ Evaluated the effectiveness of the PMT model using a random forest classifier and cross-validation techniques⬩ Aimed to mazimized accuracy and effectiveness of income qualification evaluations for more targeted social assistance programs.Tech stack: Python, pandas, scikit-learn, random forest classifier, Jupyter Notebook Show less

• 

  Lenovo

  May 2023 - Jun 2023

  Data Scientist (ML & NLP)

  Responsibilities:⬩ Supported Lenovo in understanding customer experience through Amazon product reviews⬩ Analyzed their data using syntactic processing and topic modeling approaches⬩ Normalized casings and tokenized reviews for further analysis⬩ Used parts of speech tagging to identify key nouns for subject modeling⬩ Lemmatized data to unify different forms of terms, removed stopwords and punctuation⬩ Developed a topic model using Latent Dirichlet Allocation (LDA) with 12 topics⬩ Evaluated coherence of the LDA model to ensure quality of topics⬩ Interpreted and refined topics through a business lens for actionable insights⬩ Optimized topic model by adjusting the number of topics and reassessing coherence⬩ Named topics in a business-friendly manner for client's comprehension⬩ Presented findings and interpretations in a structured table format for client reviewTech stack: Python, pandas, NLTK, Gensim, spaCy, scikit-learn, Jupyter Notebook Show less

• 

  AAL - Australian Clothing Brand

  Jun 2023 - Jun 2023

  Data Engineer

  Responsibilities:⬩ Analyzed fourth quarter sales data for AAL, a clothing brand in Australia⬩ Identified states with highest revenues through complete sales data analysis⬩ Developed targeted sales strategies for states with lower revenues⬩ Conducted detailed demographic analysis including children, women, men, and elderly⬩ Made use of effective analytical techniques to provide actionable insights on revenue generation⬩ Assessed sales trends in metropolitan areas, tier 1, and tier 2 cities⬩ Designed customized sales programs for states with lower revenues⬩ Implemented data techniques to optimize performance and address specific challenges⬩ Contributed to AAL's strategy decision making process and investment decisions⬩ Promoted a data centric culture within AAL to maximize strategic planning and growth⬩ Established a framework for future expansion and sustained success in the retail industryTech stack: Python, Pandas, Matplotlib, Jupyter Notebook Show less

• 

  Spotify

  Jun 2023 - Aug 2023

  Data Scientist

  ⬩ Executed in-depth exploratory data analysis (EDA) and cluster analysis for Spotify⬩ Improved the platform's recommendation algorithm by grouping music into discrete cohorts⬩ Utilized data collected from Spotify's API, focusing on all Rolling Stones albums⬩ Employed each song's unique ID for thorough examination of elements contributing to cohort formation⬩ Analyzed large dataset to identify patterns, correlations, and distinct characteristics influencing song categorization⬩ Established groundwork for constructing cohorts of music with comparable characteristics⬩ Conducted cluster analysis to reorganize music according to common characteristics⬩ Applied contemporary techniques to enhance Spotify's capacity to anticipate user preferences⬩ Generated song cohorts to improve personalized and engaging user experiences⬩ Gained thorough understanding of various elements influencing music groupings⬩ Provided Spotify with important observations for continuous platform improvement⬩ Ensured a more personalized and enjoyable music streaming experience for consumers Show less

• 

  K2-AI AGENCY™

  Sept 2023 - now

  I am responsible for designing and deploying AI systems. This involves developing AI architectures that align with our clients' strategic goals, guaranteeing these solutions are scalable, robust, and sustainable. I lead our technical team in executing AI projects and assuring that our solutions are delivered on time and meet high-quality standards. I work closely with clients to understand their challenges and present AI strategies that address their needs, all while aquiring knowledge on the latest and practical AI technology advancements to incorporate the most disruptive techniques into our solutions. My focus on problem solving and digital transformation, combined with my ability to effectively communicate expert technical concepts to non-technical stakeholders, guarantees that we deliver exceptional value and drive long term success for our clients. Show less

  • Founder & Chief AI Architect

    Sept 2023 - now

  • Machine Learning Engineer

    Sept 2023 - now