At TMA Solutions, I developed and optimized an LLM-powered academic advisory chatbot. I integrated LangChain and Google Gemini, and implemented semantic caching that reduced LLM API calls by 40% and average response latency from 6.1 seconds to 3.1 seconds.
I also built parallel execution pipelines that improved request processing throughput by 35%. I refactored backend components and resolved integration issues.
On Orbitama, my agentic AI marketing automation SaaS, I architected and developed a platform for content publishing, customer discovery, and customer engagement. I built a natural-language LangGraph agent that lets users execute multi-step workflows through conversational commands.
For my fruit identification project, I built a deep learning application for food quality inspection and a Flask web app to serve the model. Fine-tuned ResNet-50 transfer learning achieved 94.6% accuracy and a 93.5% F1 score.
