Harsh Bhatt

* Finetuned & Tested Diffusion models with techniques like Dreambooth, LoRA to create a personalize models for users

* Implemented a SDXL fine-tuning approach in order to achieve a 40% increase in user save rates to improve customer satisfaction

* Managed, MLOps for finetuning Stable Diffusion models. Built parallel and batch inference in order to make the generation process faster by 2x.

* Built an React App for Evaluating Different Stable diffusion models for AB Testing.

* Contributed in Image & Headshots Evaluation pipeline:

(a) Trained AI classifier to detect Ai images

(b) Built glare detector to identify glare on glasses

(c) Researched new image similarity metric

* worked with senior engineers in designing and testing TTS & voice cloning system.

* Tested & evaluated different TTS models like Tortoise-tts, TransformerTTS, etc.

* Contributed on Paper Implementation of Nancy++