Ajitesh Gupta

- Leading and mentoring the ML team in product identification, attribute classification, content generation, and automated quality checks, helping get company-wide recognition for team members - Enhanced on-site product search experience for customers using few-shot prompting and fine-tuning of MLLM...

- Leading a cross-functional team, of ML Engineers, Domain Experts, Fullstack Engineers, for producing Customer centric Content - Implementing LLM and MLLM based inference pipelines to generate content at scale - Building an application to help copywriters produce and manage content at scale - SEO o...

- Led the implementation of new model training library and infrastructure based on Huggingface and AzureML - Achieved 10x speedup in model training while having a 5x cost reduction and 5-10% improvement in F1 scores across a large collection of models - Researched and implemented customized SOTA mul...

- Led a cross-functional team of ML Engineers, Domain Experts along with Backend and Frontend Engineers to build and serve real-time recommendations - Designed and led the development of a multi-stage recommendation system with retrieval and ranking - Feature engineering and analysis for various ran...

- Prototyped various ML Recommendation algorithms - Analysed data to extract insights and develop recommendation strategies - Developed algorithms and pipelines to generate personalized shopper profiles and shopper segments - Developed ElasticSearch based real-time recommendation serving stack - Con...

- Researched and implemented various SOTA image classification models - Implemented SQL based data extraction pipeline which built and executed queries on demand based on various experiment parameters - Led various POVs which helped the company land multiple customers and generate millions of dollar...

- Designed and developed PyTorch based model library to provide 2x speedup for training - Designed and developed data and model artifact management system on AWS - Implemented SOTA models, optimizers, learning rate schedulers, loss functions for 5-10% F1 improvement across various models - Implement...