Shashikiran Alloli

Worked for the team responsible for developing a recommendation system for a home improvement company
Analyzed author and co-author network using publications data to find top influencers across the globe in different practices

Had been part of Object detection model development for store bots to detect objects on the brick-and-mortar store shelves for inventory purposes; employed CNN achieving an accuracy of 77%
Analyzed sentiments of users based on unstructured data from various data sources - eCommerce website, Reddit, Twitter
Data Preprocessing, Statistical Model, Data Visualization, Exploratory Analysis, and Hypothesis creation

Performed anomaly detection to find the dip in signals to locate the faulty cell sites, that eventually brought down the manual intervention costs by 30%, employed ML (time-series analysis) achieving an accuracy of 84%
Modeled topics of posts and reviews from Twitter, Reddit, news channel for a set of predefined tags, for a microchip manufacturer. NLP - TopicModelling
Building end to end Production ML Pipelines/ Workflow for automation of ML tasks using Sklearn, Apache Airflow TF Serving
Collaborating with a team of 5 members including data scientists and data engineers

Modeled a hybrid approach to customers' expected lifetime value (CLTV) prediction for a tech giant using both ML (random forest regressor) and DL (tensorflow2, keras) achieving RMSE - 45
Segmented customers based on their transactional behavior and also created the model to find the propensity of customers churning in the next 3 months, for a leading financial services bank in UAE with the help of ML (XGBoost classifier) with precision-recall curve 0.85, specificity 0.81, and sensitivity 0.90
Implemented cross-selling and up-selling (product recommendation) model for a financial services bank with success rate of conversion - 75%; using machine learning models
Deployed face detection and recognition for a multi-national company's annual/mid-year events and newsletters; using face_recognition through 128D face embeddings achieving an accuracy of 86%
Table detection using Deep Learning to extract financial information from Images and perform prediction; using YOLO-v4, SSD, LabelImg, OpenCV, and trained it on 1600 self-labeled images (accuracy of 78%)
Created a pipeline to translate conversational speech to text for a multinational microchip manufacturer using deep-speech3, speaker diarization (accuracy 82%), and Azure cognitive services (90% accuracy)
Guiding my team of 9 members including Data Scientists, ETL, and Data Analysts

Modeled a hybrid approach to customers' expected lifetime value (CLTV) prediction for a tech giant using both ML (random forest regressor) and DL (tensorflow2, keras) achieving RMSE - 45
Segmented customers based on their transactional behavior and also created the model to find the propensity of customers churning in the next 3 months, for a leading financial services bank in UAE with the help of ML (XGBoost classifier) with precision-recall curve 0.85, specificity 0.81, and sensitivity 0.90
Implemented cross-selling and up-selling (product recommendation) model for a financial services bank with success rate of conversion - 75%; using machine learning models
Deployed face detection and recognition for a multi-national company's annual/mid-year events and newsletters; using face_recognition through 128D face embeddings achieving an accuracy of 86%
Table detection using Deep Learning to extract financial information from Images and perform prediction; using YOLO-v4, SSD, LabelImg, OpenCV, and trained it on 1600 self-labeled images (accuracy of 78%)
Created a pipeline to translate conversational speech to text for a multinational microchip manufacturer using deep-speech3, speaker diarization (accuracy 82%), and Azure cognitive services (90% accuracy)
Guiding my team of 9 members including Data Scientists, ETL, and Data Analysts.

Performed anomaly detection to find the dip in signals to locate the faulty cell sites, that eventually brought down the manual intervention costs by 30%, employed ML (time-series analysis) achieving an accuracy of 84%
Modeled topics of posts and reviews from Twitter, Reddit, news channel for a set of predefined tags, for a microchip manufacturer. NLP - TopicModelling.
Building end to end Production ML Pipelines/ Workflow for automation of ML tasks using Sklearn, Apache Airflow TF Serving.
Collaborating with a team of 5 members including data scientists and data engineers

Had been part of Object detection model development for store bots to detect objects on the brick-and-mortar store shelves for inventory purposes; employed CNN achieving an accuracy of 77%
Analyzed sentiments of users based on unstructured data from various data sources - eCommerce website, Reddit, Twitter.
Data Preprocessing, Statistical Model, Data Visualisation, Exploratory Analysis, and Hypothesis creation.

Worked for the team responsible for developing a recommendation system for a home improvement company.
Analyzed author and co-author network using publications data to find top influencers across the globe in different practices.