 Co-founded Versos.ai, leading the development of innovative text-to-3D world generation in Unreal Engine, transforming the content creation landscape. Engineered an Unreal Engine plugin that enables single-click search and import of over 1 million 3D models, integrating multimodal language models for intelligent query suggestions tailored to scene context. Developed a state-of-the-art search tool utilizing multimodal language models, facilitating high-accuracy and rapid retrieval of images, videos, and 3D models through natural language queries. Deployed scalable APIs with Hugging Face endpoints for language model hosting, generating embeddings stored in Supabase for efficient vector space management. Built an end-to-end Retrieval-Augmented Generation (RAG) application that allows patients to interact with their medical records in a GraphDB. By converting natural language queries into Cypher queries and using embeddings for data retrieval, the application helps patients better understand their records.

 Successfully trained and deployed PyTorch implementation of yoloV5 object detection model for media content moderation to identify smoking and drinking scenes.  Analyzed the audio by computing the Short-time Fourier transform, and predicted emotions, objects, and slow motions in TV shows to help promo makers.  Developed a CNN model to identify technical cues, such as color bars, slates, scenes, and show ad break bumpers, in videos using TensorFlow.  Extracted keywords using YAKE and MultipartiteRank from the speech-to-text output for indexing the media assets in elastic search. This improved the search experience in the tool. Worked closely with clients and product managers to identify business problems and come up with ML solutions

 Worked on automating medical transcription using AI. Used DeepSpeech, an open-source model and fine-tuned it for this purpose. Reducing the time of the manual process to one-third.  Compared various speech-to-text services in AWS, GCP, and Azure and calculated the word error rate to pick the best-performing service.  Worked on image classification problem for patient segment by using the photo of the individual. Used PyTorch and Fastai to get the job done. Worked on extracting data from various social media and storing it in HDFS. Led a team of 3 junior data scientists. Also took care of the GCP admin part.  Worked on Kubeflow for end-to-end model training, tuning, and deployment.

 Worked on a finance and accounting project to develop, and implement a web-based analytics tool for solving payment-related issues for many clients including Ericsson, British Petroleum, TNT, and Nokia.  Worked in SQL Server Management Studio to play around with live production data and deploy the change and enhancement requests in prod.  Worked on analyzing, and visualizing data and used PowerBI to build a dashboard