SIRISHA PUNNAMRAJU

User Engagement:

• An application which enhances the user experience by applying AI in every step of the process and engages the user through gamification of the app. Users can enter their data in 4 different ways. They can speak to the app, upload an image or a file, enter the free text or select from a given set of choices

Responsibilities:

• Designed & implemented the ML pipeline flow
• Identified different data sources & collected appropriate data
• Cleaned & Pre-processed the data
• Prepared the Data dictionaries to train multiple models
• Evaluated the trained models and select the best model that fits the current requirement

Text Recommendation:

• Display the user a set of matching terms while they are typing the text. The search of matching terms is not just a lexical search, but a contextual semantic search based on a set of dictionary terms
• Tools & AI Models: Sentence BERT Transformer Models, Fast-AutoComplete, Python

Speech To Text Conversion:

• Collected and Prepared the audio and text data to train the speech model
• Custom trained Microsoft speech to text service for domain specific terms.
• Tools & AI Models: Microsoft Custom Speech service, FastAPI, Python

Image to Text Conversion:

• Performed Image Data Collection and Data Augmentation to test and evaluate the selected cloud Text Extract services
• Integrated the text extract service into the application flow
• Tools & AI Models: Google Vision API, AWS Textract, Spark Vision NLP, FastAPI, Python

Custom Named Entity Recognition:

• Trained and Integrated an NER model to extract entities from text.
• Tools & AI Models: RASA & Clinical NER Model

Data/Application Migration with ETL:

• Configurable Data Migration Service to migrate data from disparate sources to CTMS
• Implemented Config driven approach to pre-process, transform and map data into target system.
• Tools & Frameworks: Python & AWS

Staff Utilization & Revenue Optimization:

• Identified and Analyzed various database tables to get the staff schedules and the appointment durations
• Performed EDA in Tableau to find key insights into the data
• Calculated several metrics to assess current business performance
• Compared actual appointment hours vs staff scheduled hours to identify the staff performance
• Estimated the actual revenue vs increased revenue with the optimum staff utilization
• Tools & Frameworks: Tableau, Python for data-preprocessing

Campaign Analysis:

• Performed EDA in Tableau and provided key insights to assess Campaign effectiveness
• Derived the conversion rate for each site to estimate the number of users to contact based on several site metrics.
• Tools & Frameworks: Tableau, Python for data-preprocessing

Provided Deep Learning AI solutions to various NLP tasks involved in the application flow. Worked well in a team setting, providing support and guidance.

MLOPS pipelines:

• Orchestrated custom MLOPS pipelines for Deep learning/Transformer models
• Tools & AI Models: AZURE MLOPS, Transformer Models

Speaker recognition & Sentence Type Detection:

• Collected and prepared the data to identify the speaker and sentence type from the text
• Labeled the sentences for multi class classification
• Fine-tuned and deployed the models for better accuracy and performance
• Tools & Models: Pytorch Framework, BERT Transformer Models

Custom Named Entity Recognition:

• Annotated Data to train an NER model from scratch
• Trained an Entity Linker model to customize the NER output to deliver required set of dictionary codes for the domain specific terms
• Extracted domain specific entities based on the context
• Tools & Models: Prodigy (for Annotating the dataset) & Spacy NER Models

Semantic Word Similarity Search:

• Identify the most suitable High-level terms for a given term and associate the output with the relevant ICD dictionary codes
• Tools & AI Models: Python, Sentence Transformer Models (SBERT)

• Real Time Spark Streaming Application to migrate data from Azure SQL Server Database to AWS REDSHIFT
• Worked on Smart Patient Data Processing Platform (PDP) which analyses the patient spend patterns and predicts the probability of a patient availing the services of a hospital
• This includes collecting and storing patient data into HDFS and processing using SPARK core and SPARK sql modules in Python (Pyspark)
• Finally developing a predictive model using spark MLlib
• Developed many use cases and mini projects on big data such as
• Uber dataset analysis using spark sql (in pyspark)
• 911emergency helpline data analysis using spark core (in pyspark)

• Project: Bank Marketing Campaign Analysis using R and Tableau
• To develop the best classification model to predict the potential clients for the Term Deposit Subscriptions that are being offered by the bank
• Thoroughly implemented Exploratory Data Analysis Techniques to gain insights into the data
• Developed 5 different classification models such as
• Logistic Regression,
• SVM,
• Decision Trees,
• KNN and
• Random Forest
• Determined the best classification model for the current data based on the performance statistics such as prediction accuracy, AUC and ROC curves
• Academic Projects on Machine Learning and Data Analytics:
• Audit Analytics in python: Evaluating Audit opinions to predict the company performance
• Loan Default Prediction, Stock Market and Portfolio Management in R