Zeerak Nadeem Wyne

Experience

Sr. Data Scientist

Apr, 2021 - Present

Untangl

- Lead a team of 4 to design, create, and deploy data analytics and machine learning pipelines.
- Implemented automated data processes, services and tools for ETL, ingestion, and EDA report generation.
- Performing Ad-hoc Analysis, Root Cause Analysis, and Clustering Analysis to produce comprehensive reports.
- Generating insights from analysis and communicating results to stakeholders.
- Created Power BI and Tableau dashboards for analysing performance of call center KPIs.
- Collaborated with teams to design and deploy serverless architectures for Data pipelines.

Achievements
- Agent Segmentation Analysis using K-means clustering to understand specific skills used/not used by agents in calls that drive sales; Increased average agent conversion by 5% and sales by 20%.

- Agent Attrition and Behavioural Analysis to get visibility on reasons impacting attrition and designed a Tableau dashboard for monitoring; Decreased average agent attrition for clients by ~8%.

- Developed Agent Skill Importance algorithm to suggest agent skills for customer satisfaction: increased customer satisfaction scores by 15%.

- Created call center KPI dashboard which reduced the analysis time by 40%.

Key Projects
- Named Entity Recognition (NER): Developed the NER pipeline to identify 2000 different tags related to Speech, Customer Experience, and various call center KPIs in the transcriptions. This enabled us to use these tags to perform KPI calculations and generate insights for a key dash boarding service we offer the clients.

- Audio Speech Transcription (ASR) pipeline: Deployed ASR on Kubernetes to generate transcriptions of call recordings for clients. Designed the data transformation and ingestion pipeline that processed the generated transcripts and stored them to an Elasticsearch database. Used machine learning models to generate call Emotions, Sentiment, Summarisation, and De-identification of transcripts.

- Context Engine API: This was the API developed to use the tags generated by NER to identify Agent Skills, Call Elements, Call Phrases and Signals for Sales, customer experience, Risk from the transcripts. Singlehandedly developed the calculation logics and API to identify these elements, which were the backbone of analytical dashboards we were offering to the clients.
Technologies: Python, AWS, Elastic search, Firebase, Machine Learning, Tableau, NumPy, Pandas, PostgreSQL, ETL, GCP, Data Engineering, scikit-learn, noSQL, Data Analysis, Power BI, Git, Business Analyst, Business Intelligence

Data Scientist

Apr, 2020 - Apr, 2021

STARZPLAY

- Performed various analysis for ad-hoc reporting and prepared decks for stakeholders and potential investors.
- Cut projected time for data analysis and reporting by one week using automation through Python scripting.
- Created and automated tableau dashboards for companywide KPIs and campaign performance metrics.
- Supervised 5 different projects: Customer churn prediction, Customer lifetime value estimate, Customer Segmentation Analysis, Data Scraping/Enrichment pipeline, and Content Recommendation Engine.

Key Projects
- Customer Segmentation Analysis: Performed clustering analysis using K-means and Agglomerative clustering to identify characteristics of paying and churned users from 6+ acquisition channels for targeted campaigns and STARZPLAY content acquisition; decreased monthly churn by ~5% and increased sign-ups by 10%.

- Data Scrapping pipeline: Designed and deployed the scrapping pipeline to scrape data for existing and new content on the platform. This data helped us perform consumption analysis to surface trends in user consumption behavior

- Customer Churn Analysis and Prediction: Performed churn analysis to identify potential reasons of customer churn and performed feature engineering to develop a churn prediction model to identify customer likely to churn using Decision Tree Classifier to perform targeting marketing; decreased churn by 3%.

- Content Recommendation Engine: Developed the building blocks of the recommendation engine to learn from data and recommend TV shows and movies to users based on their interests and content information using collaborative filtering.
Technologies: Data Analysis, Machine Learning, Python, Excel, Amazon Redshift, Tableau, Selenium, scikit-learn, Scrapy, SQL, Matplotlib, Seaborn

Data Analyst / Machine Learning Engineer

Oct, 2019 - Present

Fiverr

- Delivered custom solutions to 50+ clients in domains of Data Science, ML-Ops, Machine Learning, Computer Vision,
Forecasting, ML Model deployment, ETL pipelines, Data Analysis, and Visualizations.
- Achieved a Level 1 seller badge for customer satisfaction and maintaining an overall rating of 5/5.
Technologies: AWS, Python, Pandas, Data Science, Machine Learning, Data Analysis, Deep Learning, OpenCV, Matplotlib, Seaborn

Deep Learning Engineer

Aug, 2018 - Mar, 2020

OnStack

- Developed POCs for ML and IoT use cases in domains of Video Analytics and Predictive Analytics.
- Developed APIs in collaboration with the Front-end and DevOps teams for ML Projects.
- Researched, fine-tuned, and optimised deep learning models for edge deployment (Raspberry Pi, Jetson Nano)
- Worked on multiple deep learning projects in parallel to deliver timely work to clients.
- Researched on methodologies to prepare custom solutions for clients.
- Trained and fine-tuned custom neural network models for Computer Vision and Time Series forecasting

Key Projects
- Smart Energy Monitoring: Developed and deployed the energy consumption forecasting algorithm for the smart meters to forecast weekly and monthly energy consumption and predict the monthly electricity bill of a household. Tried various models like Prophet, XGBoost, and LSTM. Achieved a Mean Absolute Error of 145 on LSTM and deployed as an API.

- AI Office Assistant: Developed a virtual assistant to interact with employees and visitors in the office. Functions of the assistant included RASA Chatbot, Voice Feedback, Employee Identification, Form Filling and Meeting scheduling.

- Facial Recognition Attendance: Developed a custom facial attendance solution for KPMG to identify employees and mark their check-in/check-out time and grant access. Trained a custom Siamese network for KPMG employees and deployed the system on an edge device to be installed in the company.
Technologies: Machine Learning, Deep Learning, Python, API Development, Flask, Chatbot, Raspberry Pi, Keras, TensorFlow, OpenCV, Matplotlib, Seaborn, Business Intelligence, Business Analyst