My Impact

















Commercial effectivity
With client retention just over 80%, there was a commercial need to grow while maintaining the existing client representative officers (CROs) number. This presented a problem that required CROs to be more efficient. The project involved effective use of AI & strong collaboration with implementation executives to discover the chances of clients not renewing their products, the reasons for such attrition and the best courses of action. This resulted in reducing the attrition rate by more than 60% within 2 years, driving ~20% YoY growth and saving over £4m annually
The business has a long-standing history of marketing budget allocation based on intuition. Their website appeared to lack optimisation due to high bounce rates. To improve these, incremental mix market modelling was used to optimise budget allocation and multi-touch attributional modelling was used to improve site performance and measure campaign effectiveness. The initial state distribution offered insights into landing pages. Page attribution measured campaign effectiveness and page quality, helping with bounce rate problems. Because of the increased dynamism of the market environment, the market mix model was automated so incremental updates were based on subsequent stability and/or characteristic analysis.  This drove Increased campaign monthly ROMI by ~12%, a significant reduction of bounce rate, and growth in conversion rate (~17%) on the website.
 
Business operational efficiency
Reporting within a vertical organizational structure could be challenging. This project enabled efficiency within attendance reporting thereby creating a scalable solution that converted free text business job titles into one of the five seniority levels. This saved the organisation an equivalent of 2 business analysts.
Adopted a multi-stage approach in the introduction of a cloud-based speech analytics solution with Call Miner and webchat analytics. Deployed advanced text analytic solutions to classify calls into established call topics to understand drivers of customer actions reporting outcomes on L0 BI platforms. This mining analysed almost 10m calls annually and delivered £m in operational cost savings
 
Data capability efficiency
Customer data were siloed and hosted in differing solutions both on cloud-based and on-prem. This significantly delayed delivery of projects, the quality of data, and the scalability of solutions. In response, created a cloud-based client 360 database enabling the availability of high-quality data for scalable rapid development of AI solutions for the business. This reduced time to delivery of data-driven solutions by more than 40%.
The importance of ensuring the model performance in production is as high as when it is being developed can not be overstated. A 2% drop in the accuracy of any model could result in organisation losing £’m annual. This project involved the development of an MLOps solution dashboard providing real-time monitoring for over 50 productions models in areas of prediction accuracy, the business value generated, data quality, and data structure. This saved the organisation from employing at least 4 FT analysts.
Innovation
Failed engineer installations were costing the organisation £millions and there was a risk of regulatory sanctions. With a goal of meeting the regulatory target, an AI-based system was developed that highlighted the drivers of installation failures at a strategic level, identified the customers who were likely to have failed installations, and recommended actions to mitigate chances of failure. The effective use of this system led to a 9% improvement in successful installation rate, saving c£1.5m monthly
The school had a history of underperforming both locally and nationally. Charged with a goal of bringing performance within the national average (0.67 grades lower), a tool was developed that predicted students’ end of schooling grades based on their standard level performances and demographic information. The tool clustered students of similar strength profiles into learner groups enabling targeted interventions and suggesting the most effective topics to be taught to each learner group. This tool enabled 24% YoY improvements in the proportion of students achieving the gold standard. The school performance exceeded those of national levels by 0.26 grades.

Tooling and Technical Expertise

Data Technology: AWS, Azure and GCP
Big Data & Database Systems: RDMS, Spark, Hadoop, SQL, NoSQL, HQL, Data architecture, modelling and optimisation
Data Science: Statistics, Machine Learning, Natural Language Processing, Recommender Systems, Deep Learning, XGBoost, Recurrent Neural Network, Artificial Neural Network, GIT, Latent (Hidden) Markov Models, Classification, Reinforcement Learning, Clustering, Regression, Stimulation, Optimisation, Convolutional Neural Network, Seq2Seq Modelling, Topic modelling (LDA)
Tech Tools: Spark ML, Keras, Tensorflow, IBM SPSS Statistics, MPlus, Microsoft Excel, Apache Storm, Hive, Ambari, Kafka, Microsoft Azure Suite, Tableau, Power BI, Pega, Unica, CI/CD
Programming Languages: Python, Scala, SQL, R, Mplus
Libraries: Pandas, Statsmodels, NumPy, SciPy, PyTouch, Scikit-Learn, Matplotlib, Spark MLlib, Plotly, Folium, Tensorflow, Keras

Search

Gallery

My Impact

Commercial effectivity

Business operational efficiency

Data capability efficiency

Innovation

Tooling and Technical Expertise