Sujitha Ravichandran

Generative AI Engineer/Researcher

AICEBERG (May 2024 - Present)

• Enhanced Large Language Models (LLMs) such as Llama, Mistral, and Falcon, achieving a 25% improvement in performance and efficiency. Employed GPU-enabled machines with PyTorch and CUDA libraries, reducing training time by 40%.
• ADesigned and managed ETL pipelines for complex datasets, with a focus on instruction-based fine-tuning of LLMs. Boosted data scalability and reliability through comprehensive data processing and integration strategies.
• Handled datasets from sources like English Wikipedia, improving embedding utility for text similarity and analytics. Utilized custom Python scripts and Spark to enhance data extraction, cleaning, and manipulation, achieving a 50% efficiency increase.
• Leveraged AWS services such as Redshift and S3 for data warehousing, ensuring seamless integration with existing infrastructure. Improved data access and analysis capabilities for global teams, cutting data retrieval time by 35%.
• Conducted prompt analysis to assess the relevance of user inputs to customer data. Applied Retrieval-Augmented Generation (RAG) techniques to provide context-based responses, increasing response accuracy by 20%.
• Managed customer data onboarding using LangChain custom loaders, processing data in various formats, chunking it, and adding metadata. Fine-tuned LLMs on customer data using AWS S3, improving processing speed by 30%.

Data Processing, Analytics, and Data Visualization Specialist

MATA INVENTIVE (Aug 2023 - Dec 2023)

• Collaborated with cross-functional teams to understand business requirements and extract data-driven insights for decision-making.
• Automated data transformation tasks using Python scripts, reducing manual workloads, and increasing data processing efficiency and achieved a 30% reduction in processing time, leading to faster decision-making.
• Conducted A/B testing and statistical analysis to measure the impact of data-driven decisions on key performance indicators (KPIs). Improved KPIs by 15% through iterative testing and optimization, contributing to data-backed strategic decisions.
• Created data visualizations with PHP, highlighting key insights in machine monitoring and inventory data.

Graduate Research Assistant

UNIVERSITY OF HOUSTON (Jan 2023 - Jul 2023)

• Deployed and fine-tuned Transformer-based language models from the Hugging Face library for image-to-image translation tasks in the field of computer vision. Achieved a 15% improvement in translation accuracy compared to baseline models.
• Analyzed the architecture details of different Transformer models, including variations such as BERT, GPT-2, and RoBERTa, and conducted layer-by-layer analysis.
• Reduced model training time by 20% through optimization of attention mechanisms, enhancing overall efficiency.
• Extended the application of Transformer models beyond image-to-image translation to natural language tasks, this involved fine- tuning models for tasks such as text classification, sentiment analysis, and language generation.

Machine Learning Engineer Intern

INDIAN INSTITUTE OF TECHNOLOGY HYDERABAD (Jul 2021 - Aug 2022)

• Attained fine-tuned state-of-the-art object detection models to specifically detect faces within datasets. Attained an impressive accuracy rate of 78% in detecting faces.
• Played a crucial role in a team that developed a new architecture for image dehazing, a significant achievement that has been recognized through a publication.

Data Science Intern

NATIONAL INSTITUTE OF TECHNOLOGY TIRUCHIRAPPALLI (May 2021 - July 2021)

• Employed both supervised and unsupervised machine learning techniques to analyze and work on ontologies datasets.
• Utilized RDF (Resource Description Framework) and OWL(Web Ontology Language) for constructing comprehensive ontologies in Extractive Metallurgy, leveraging their capabilities for complex data representation and interrelation. Employed Protégé for efficient ontology editing and integrated machine learning models for automation and enrichment of the ontological data.
• Contributed to a research paper submitted to the International Conference on Data Science and Security, demonstrating proficiency in data science methodologies.