Naresh Joshi

Founded and led Techsergy, building a video analysis platform using small language models to process 500 petabytes of video, with 1 hour of video processed in under 30 seconds. Developed vision language models, reducing processing time by 95% through quantization, pruning, distillation, and fine-tuning.

Led AI research at Tether, building a video analysis platform processing 500 petabytes of data and compressing video models to 8B parameters while maintaining quality of 235B MoE. Implemented event-driven architecture processing around 5000K videos per week.

Built an LLM-powered intelligent web crawling system using Named Entity Recognition, reducing maintenance costs by 80%. Developed and optimized Small Language Models for on-device NER, enabling efficient edge inference.

Designed and implemented services using CNN and LLM-based models for document scanning and entity extraction from legal contracts, digitizing over 100 million contracts. Restructured application architecture, achieving a 70% increase in digitization speed and 60% reduction in costs.

Developed an inventory management system for Apple for real-time monitoring of over 3000 servers globally. Built services for CI/CD pipelines and a central repository tracking over 2000 stories, and engineered LEX platforms for Infosys enabling training for over 300K users per day.

Bachelor of Technology in Mechanical and Automation Engineering from Indraprastha University, New Delhi, from 2013 to 2017.

Machine Learning Specialisation completed through Coursera, offered by Stanford University and Deep Learning AI.

Programming in C course completed through NPTEL, offered by IIT Kanpur.

AWS Cloud Practitioner certification obtained from Amazon Web Services.