Bala Chandra Sekhar Reddy Bhavanam

Proven expertise in software development and problem-solving, honed at CGI through advanced data analysis and visualization using Python and SQL. Excelled in agile environments, demonstrating a keen ability to translate complex data into actionable insights. Skilled in both technical and collaborative capacities, significantly enhancing project outcomes.

E-commerce Customer Experience and Market Intelligence, Designed and developed a Customer Data Analytics Platform to enhance customer experience and market understanding for E-commerce businesses. Identified and addressed challenges in utilizing customer and market data, enabling better decision-making and personalized customer experiences. Integrated machine learning frameworks such as TensorFlow and Scikit-Learn to analyze data patterns, adapt to new trends, and generate actionable insights. Implemented real-time data visualization tools like Tableau, enabling comprehensive reporting and monitoring of market trends with a user-friendly interface. Provided a holistic solution combining data management, advanced analytics, and visualization, helping E-commerce companies outperform competitors and improve customer satisfaction. Documented the project lifecycle from data cleaning and exploration to model development, ensuring a structured and detailed overview of progress and outcomes. Hate Speech Detection, Developed a comprehensive social media-specific dataset with multi-class labels, annotated under well-defined rules and strong annotator agreement to ensure accuracy. Explored and implemented text mining features for hate speech detection, including character and word n-grams, dependency tuples, sentiment scores, and pronoun counts. Utilized Latent Semantic Analysis (LSA) for dimensionality reduction, enabling efficient processing of complex and non-linear data models. Applied and evaluated advanced machine learning models, with CAT Boost achieving the best performance for accurate hate speech prediction. Focused on distinguishing and predicting different categories of hate speech using baseline and self-discovered features tailored for problem suitability. Highlighted the societal importance of the project by addressing hate speech's impact on violence, societal imbalance, and human rights violations.

Abstractive Summarization is Improved by Learning Via Semantic Similarity, Developed a semantic similarity-based training technique to enhance abstractive summarization, leveraging pre-training and fine-tuning phases with the BERT model. Proposed and implemented an additional layer to compute semantic similarity scores between generated and reference summaries, improving alignment with human evaluation standards. Achieved state-of-the-art performance with a ROUGE-L score of 41.5 on the CNN/Daily Mail dataset. Conducted human evaluations to supplement automated validation, resulting in superior performance compared to baseline models and reference summaries. Addressed challenges in abstractive summarization by effectively handling multiple valid predictions and improving the semantic coherence of generated summaries., https://doi.org/10.1007/978-981-19-6631-6_23