About Me
Data Scientist specializing in deploying Generative AI and Retrieval-Augmented Generation (RAG) solutions using AWS Bedrock and SageMaker.
Skilled in building scalable AI solutions with fine-tuned foundational models and integrating knowledge bases for various industry applications.
Proficient in developing and deploying machine learning solutions using AWS SageMaker and AWS Bedrock, with hands-on experience in data engineering and advanced analytics to drive decision-making processes across various business functions.
Certified in AWS Machine Learning Specialty, AWS Machine Learning Associate, and AWS AI Practitioner; Google Cloud Certified ML Professional; and Azure AI-102: Designing and Implementing Microsoft Azure AI Solutions.
Demonstrated ability to design and implement data pipelines and robust AI solutions, including generative models and cloud-based data management systems, to improve operational efficiency and data quality.
Designed and implemented scalable architectures tailored to client needs using AWS AI and Generative AI services, ensuring alignment with business objectives and industry best practices.
Proficient in deploying and monitoring AI/ML models using platforms like AWS, GCP, and Azure, ensuring scalability and reliability in production environments.
Skills
Data Science & AI
ML Frameworks
Data Wrangling & Viz
Cloud Platforms
AWS Services
Experience
Data Scientist
- Engineered and optimized RAG systems, enhancing data retrieval pipelines and significantly improving efficiency and accuracy of information extraction.
- Developed a scalable AI-driven pipeline for text extraction, classification, and PII redaction in PDF documents using AWS Textract, Rekognition, Bedrock LLM, and Comprehend, achieving high accuracy rates.
- Built a serverless React application with embedded Amazon QuickSight dashboards using AWS Lambda, API Gateway, IAM, and QuickSight Embedding SDK for anonymous access.
- Implemented fine-tuning processes on pre-trained LLMs for specific text and image processing tasks, achieving superior model performance and reliability with Amazon Bedrock Foundation Models.
- Demonstrated proficiency in using transformer architectures for generative AI applications, leveraging both large and small language models (LLMs and SLMs) with Hugging Face's Transformer library.
- Implemented Virtual Try-On and RAG-based chatbot functionalities for apparel recommendations, significantly enhancing customer experience for a fashion e-commerce client.
Data Analyst Trainee
- Designed and maintained robust data pipelines and ETL processes to support data ingestion, transformation, and storage for large-scale AI applications.
- Adept at preparing and preprocessing large datasets for training language models, including text normalization, tokenization, and augmentation.
- Skilled in creating interactive and comprehensive dashboards with tools like Tableau and Power BI, facilitating data-driven decision-making.
Data Engineer Intern
- Implemented data storage solutions, including storage accounts, database storage, and file storage.
- Designed, deployed, and managed Azure cloud solutions.
Research Engineer
- Worked on Sensor Fusion for self-driving cars, utilizing radars and cameras.
- Objective was to make radar an object detection tool.
Projects
Real-Time Insights with AWS QuickSight & Q
Implemented an end-to-end embedded analytics solution by integrating Amazon QuickSight dashboards into a React application for anonymous users. Used AWS Lambda, API Gateway, IAM, and the QuickSight Embedding SDK to securely generate and embed dashboards.
Robust & Scalable Document Processing
Developed a scalable AI-driven pipeline for text extraction, classification, and PII redaction in PDF documents. Achieved 99% classification accuracy, 99% PII masking precision, and 96% target entity extraction.
Claude-Haiku and Opensearch RAG-Based Insurance Advisor
Developed a RAG-based Insurance Advisor using Claude-Haiku, Amazon Titan Embeddings, and OpenSearch to compare policies from multiple insurers. Indexed insurer PDFs and leveraged AWS Knowledge Base Retrieve-and-Generate API to provide personalized, profile-driven recommendations based on user queries and needs.
RAG-Enhanced Chatbot for Fashion E-commerce
Built a RAG-based chatbot for a fashion e-commerce platform using Amazon Kendra to enhance product discovery through natural language queries. Enabled contextual search by passing metadata (e.g., color, occasion) to Kendra, delivering personalized recommendations like "red dresses for a party" in real-time.
Text to Video Generator
Pioneering a Text-to-Video Generator, working on Stability AI and large language models to seamlessly translate textual content into dynamic visual narratives. This ML engineering project combines cutting-edge techniques, ensuring stability and coherence, while harnessing the power of language models to generate engaging and impactful videos with precision and creativity.
Email Spam Detection
Developed an email spam detection system by fine-tuning ROBERTa on a custom customer dataset. The model accurately classifies emails as spam or not, improving filtering efficiency and reducing false positives.
Certifications
Contact Me
Feel free to reach out for collaborations, opportunities, or just to say hello!
Email: neerajpokala143@gmail.com
Phone: (+91) 7386153703
LinkedIn: linkedin.com/in/neeraj-pokala29
GitHub: github.com/Neerajpokala