Working on core GenAI infra and business use-cases. Optimizing operational & CS workflows with Multimodal LLMs

- Built NLP and Transformers based Text classification algorithms that helped US clients identify vulnerabilities in the sign up process of the top 1 million websites.- Developed a Dockerized Django Restfull api to wrap the OpenAi services for a custom-made domain-specific virtual assistant (this included the DB design, prompt engineering, and asynchrony implementatiom which helped reduce the inference time of some endpoints from 50mins to 2mins)- Designed and constructed a data pipeline that gathered, filtered, cleaned and preprocessed thousands of images using image processing techniques like face detection, histogram normalization, morphological dilation and much more. - Used the previously mentioned data to implement and train a state of the art architecture from a research paper for the face swapping task.- Led the development of two generative AI applications. One GPT-based virtual assistant for product managers and various diffusion-based image generation models for e-commerce platforms.