This repository contains few of the major projects, solutions, products, initiatives worked and led by me during my professional career. This is non-exhaustive and I have attempted to provide a very high level summary of major work performed in last few years, leveraging Data Science and Artificial Intelligence. Not all projects are captured here. Only significant Data Science focused projects are captured to provide a reflection of my deep expertise in the sector and significant global outcomes accomplished as part of these initiatives.
(Below is a high level view of major recent projects, the link reference on first column will provide details around each project)
| Project/Product/Initiative Name | Description | Data Science Techniques Used | Business Outcomes, Impact and Value |
|---|---|---|---|
| Failure prediction model for Predictive Maintenance - Click Here for Details (World's largest Bank by Assets, Leading Banks from Europe) | End to end ensemble based advanced machine learning solution to identify upcoming failure predictions for devices and components. The Data Science solution involved Microservices based cloud native solution using MLOps; GitHub pipeline for maintainability, scalability and flexibility; code development using Python, R, IBM Watson suite of products; Algorithms used – Random Forest, XGBoost, LightGBM as part of experimentation and model solutioning. | Methodology: CRISP-DM, Techniques: Classifiers, Tree based Ensemble learning, Algorithms: Random Forest, XGBoost, Logistic Regression, Model Evaluation: Precision, Recall, F1score, Deployment: IBM Watson, IBM Cloud, Microservices based architecture, MLOps framework (Model serving, Model monitoring and Model management) | Public client reference, 2.03 p.p. availability improvement within 4-6 months of implementation, #5 to #1 in availability for region (network operator rankings) within 4-6 months of implementation, USD1300 per device per year cash in transit saving, Failure rate reduced by 83% (from 1.2 to 0.2 tickets per device p.m. compared to non-serviced devices at 0.8) |
| Forecasting equipment sales - Click Here for Details (One of the Manufacturing Major) | The objective was to develop an approach to forecast sales for customer’s machinery equipment for next 3 years and customer wanted us to be able to provide them clarity with exploratory data analysis so that they have “end to end visibility” of what is happening. I have used machine learning models using ARIMA, Holt winters, LSTM to provide forecasting solutions | Data Science CRISP DM framrework, Forecasting, ARIMA, Holt Winters, LSTM, Feature Engineering | Machine Learning solution provides ability to plan equipments for 3-6 months in advance for Customer and drive their business effectively, Cost savings up to 20% on y-o-y comparision |
| Customer analytics solution - Click Here for Details (One of the Banking and Financial Services Major) | The customer which is a global financial services major was facing challenges profiling their customers / end users that are vulnerable to churn and wanted us to help them in devising retention strategies for retention of their end customers. We have used Machine Learning solution to classify customers in different buckets and help in decision making and insights. | Data Science CRISP DM framrework, Classification, Random Forest, Logistic Regression, Decision Trees, F1 score | Ability to provide Next Best Offers to Customers and Increase up to 8%-10% of their revenue |
| Clustering patients for specialty pharmacy - Click Here for Details (World's leading Biotech firm) | Clustering of patients for Specialty pharmacy for one of the largest biotech major to provide an ability to view patient services metrics, view weekly patient update based on specific product, to understand prescriber uptake and to cluster group of patients with similar characteristics to be able to offer specific product and services to those set of patients | Data Science CRISP DM framrework, PCA for dimenstionality reduction, Feature Engineering, KMeans, Euclidean distance, Jacquard's distance, Gower's dissimilarity coefficient, Elbow criteria | Customer / Patient satisfaction increase up to 10 percent points, Customer visit cost to hospitals reduced by up to 40 percent points, Alignment of drugs/products efficient due to better inventory of those products |
| Predict Insurance Renewal - Click Here for Details (World's one of the largest insurance service provider firm) | Ability to develop an approach and solution to predict the propensity of insurance renewal, to be able to create an incentive plan for agents (at policy level) to maximize net revenues from these policies | Ensemble techniques used, Algorithm such as Regression, Random Forest, Decision Tree, LightGBM were used for experimentation | Value realized within 3 months of implementing actions based on predictions shared are as follows: a) Client used the prediction probability to take better informed decisions about their end customers – who are more likely to renew the insurance, b) Computing the incentive amount for agents helped them strategize entire process and how to allocate agents accordingly and manage their incentives. Effort towards campaigns were managed better after implementation of the solution, c) Client wanted to perform this as an ongoing effort (generating prediction monthly for their business) for next few quarters to see the impact and leverage from it. |
Key dimensions:
- Outcomes, Value and Impact demonstrated to Clients, Stakeholders with tangible aspects such as - Public Client Reference of the work, % Cost savings, % Availability improvement, % Prediction accuracy etc.
- Confidence shown to Clients, Stakeholders by demonstrating end-to-end CRISP-DM (Cross Industry Standard Process for Data Mining) framework
- Recommendations as a partner in the areas of MLOps, ML Explainability, Data drift, Model drift, Responsible AI/ML, Ethical AI/ML etc.
- Multiple breadth of Data Science problems solved such as Statistical inferences, Classification, Clustering, Neural Networks, Forecasting, Survival Analytics, Deep Learning etc.
- Use cases solved for multiple domains such as Banking and Financial Services, Insurance, Retail and CPG, Healthcare, Manufacturing etc.
- Use cases solved across functions within "X Analytics" for Data Science such as - Customer Analytics, Risk Analytics, Marketing Analytics, Predictive Analytics, Healthcare Analytics etc.
- Innovation and Thought Leadership dimension linked to Data Science work (in terms of filing Patents, presenting papers in conferences, blogging, mentoring etc.)
Contributions in Data Science and AI Thought Leadership / DigiTech Sector
- June 2022: KeyNote Speaker at ICIMSI 2022 - 2nd International Conference on IMSI and Certification of Appreciation - Twitter Ref on
Value and Impact of Data Science and AI in Industry 4.0/5.0 - June 2021: Webinar on Accelerate Data Science and Responsible AI The YouTube Link
- May 2021: Panel Discussion at Kaggle Days Meetup Delhi NCR Chapter on Data Science Ecosystem and Hiring Prospects in India
Disclaimer: These are captured and documented to demonstrate my success, project details, global presence, value and impact created, etc., and are best to my knowledge and experiences. This does not represent any opinions from or on any firms or any institute.