Ranjith Maliga Guruprakash

Architected production-ready multimodal Generative AI assistant achieving 95% response accuracy with 75% faster response times. – Built containerized multimodal AI system processing text, image, speech, and audio inputs using Docker on Linux VM, enabling portable production deployment across environments – Engineered high-performance RAG (Retrieval-Augmented Generation) pipeline with AWS Bedrock and Chroma DB vector database, improving AI response relevance by 40% and reducing query latency by 35% – Optimized data retrieval architecture implementing Redis caching and AWS S3 integration, cutting response time from 3.2 seconds to 0.8 seconds (75% improvement) – Designed and deployed multi-agent workflow architecture for complex query handling, delivering context-aware solutions with 92% user satisfaction score – Collaborated cross-functionally with R&D teams to integrate multimodal AI capabilities into product roadmap, accelerating feature development cycle by 25%

Developed scalable LLM systems and multimodal AI solutions for blockchain technology company, focusing on video understanding and natural language processing applications. – Designed and built scalable LLM systems using TensorFlow and PyTorch, enhancing performance through advanced NLP techniques and hyperparameter optimization – Improved user engagement metrics by iteratively fine-tuning AI responses for optimal interaction quality and context relevance – Integrated and assessed video understanding models using Gemini and OpenAI APIs, ensuring high-quality performance across diverse content types – Developed robust data preprocessing pipelines to streamline model training workflows, reducing training time and improving model accuracy – Implemented LangChain framework for orchestrating complex AI workflows and prompt engineering optimization