Deep Dwivedi

• Resolved problems, improved operations and provided exceptional service.
• Worked well in a team setting, providing support and guidance.
• Acted as a team leader in group projects, delegating tasks and providing feedback.
• Organized and detail-oriented with a strong work ethic.
• Utilized IBM CPLEX in campaign budget allocation, One of the key aspects of sales advertisement optimization is determining how to allocate the advertising budget across different channels, such as online ads, TV, radio, or print media. CPLEX was used to formulate an optimization model that considers various factors like target audience, reach, costs, and expected returns. The objective was to maximize the overall impact or return on investment (ROI) within a given budget constraint.
• Online automated and AI-driven Proctoring system - To conduct secure online examinations we developed this system by using CNN(VGG Net and ResNet) along with OpenFace, OpenCV, Yolo-V3 and Deep Neural Networks for training on more than 50K images.
• Pre-requisite knowledge graph on syllabus ontology for knowledge gap analysis
• Slang and hateful chat/speech detection during online classes
• AI and computer vision driven subjective and objective OMR tests
• AI driven Advanced content effectiveness evaluator – AI based advance UI/UX testing method which does eye tracking , emotion analysis and also tracks physiological modalities using EEG, PPG, GSR sensors to correctly analyze the effectiveness of the displayed content.

Other Activities

o Published 6 patents and 6 more under processing

o 3 Research Paper in Core A conferences AAAI, AIED 22,LAK24

o K-12BERT: BERT for K-12 education - https://arxiv.org/abs/2205.12335

o Obj2Sub: Unsupervised Conversion of Objective to Subjective Questions

https://arxiv.org/abs/2206.11848

o Effecti-Net: A Multimodal Framework and Database for Educational Content Effectiveness Analysis - LAK24 - submitted camera ready

o Published an article at CXOtodaymagazine on how AI and Online Education helping our environment by reducing carbon foot prints.

o https://www.cxotoday.com/news-analysis/world-environment-day-2022-technology-companies-onprioritizing-

sustainability-and-driving-transformative-environmental-change/

o Gave Talks on Data Science and how to build a successful career in this field

o https://techcareershow.transistor.fm/episodes/e3-deep-dwivedi-long-term-plan-for-your-datascience-career

• Worked with stakeholders to develop quarterly roadmaps based on impact, effort and test coordinations.
• Utilized advanced querying, visualization and analytics tools to analyze and process complex data sets.
• Extracted and assessed data from databases to drive improvement of product development and business strategies and processes.
• Applied statistical and algebraic techniques to interpret key points from gathered data.
• Ad Placement and Scheduling: CPLEX was used to determine the optimal placement and scheduling of advertisements. We considered factors such as ad placement costs, audience demographics, time slots, and competition. By formulating an optimization problem, CPLEX was able determine the best combination of ad placements and schedules to maximize reach, visibility, and impact within the given constraints.
• Death of Interest Model - We have been facing the problem of stale and repeated recommendations since very long, this project highly focuses on removing stale recommendation on the basis of the user response model, which has been trained upon the user response patterns towards songs for more than 3 months data. Prediction for probability if a user will click on the song gets computed by analyzing the past 7 days response towards the song. XGBoostmodel is used as a training model. After deployment this has increased CTR of the Recommendation rail by 3.5%.
• Dealt with heavy volumes of Data approx 100TBs and trained model in distributed fashion using AWS EC2 clusters with Spark.

• Worked with stakeholders to develop quarterly roadmaps based on impact, effort and test coordinations.
• Applied statistical and algebraic techniques to interpret key points from gathered data.
• Waste optimisation using CPLEX, in this project we first of all did the spectral analysis of smoke and optimized raw materials in order to reduce toxic wastes.
• Vessel Routing Problem Successfully addressed complex logistics challenges by implementing cutting-edge solutions for the Vehicle Routing Problem (VRP).
• Utilized Python, Gurobi, PuLP, and CPLEX to design and optimize routing algorithms.
• Developed efficient and scalable algorithms to minimize transportation costs, reduce delivery times, and improve resource allocation.
• Collaborated closely with cross-functional teams to understand client requirements and customize VRP solutions for specific needs.
• Conducted extensive testing and performance optimization, achieving a [specific metric improvement] in route efficiency.
• Regularly maintained and updated VRP algorithms to adapt to changing business demands and market dynamics.
• Documented code, algorithms, and optimization strategies for knowledge sharing and future reference.

Major American Retailer client: Team lead for algorithmic development of Human Detection and

Tracking in video analytics domain using various digital image processing and machine learning concepts,

also worked with AWS platform for scaling up the computational architecture.

• Major Bank client: Exploratory data analysis on large datasets to arrive at insights and cater to ad hoc

client requests (credit worthiness, loyalty towards bank) and provide potential problem solutions.

• Innovation & Development Team: Published white paper for the exploration of Probabilistic

Programming in the field of Machine vision and Algorithmic Trading.Team lead for the development of

Bayesian cointegration for stock market Time Series data.