Jr. Ai Engineer
CurrentDeveloped Affective Computing aspects for robots- Developed a python version of the COVAREP MATLAB package for emotion analysis in speech. - Researched visual question answering(VQA) tasks using vision language models and Large Multimodal Models. Familiar models: Closed source: Claude-3 (Sonnet, Opus, Haiku), GPT-4( 4o, 4o mini) , Gemini-1.5(Pro, Flash).Open source models: LLaVA (1.5, 1.6), MiniGPT-4, Idefics-1, Idefics-2, LLaVAPhi-2, Phi-3-Vision - Integrated the above mentioned models in a web applicaiton for a comparaitive analysis, with the help of streamlit and FastAPI. Optimized the pipeline with asynchronus data streaming for large outputs, reducing latency.