Anant Gupta
Development
Uttar Pradesh, India
Skills
Machine Learning (ML)
About
ANANT GUPTA's skills align with IT R&D Professionals (Information and Communication Technology). ANANT also has skills associated with Programmers (Information and Communication Technology). ANANT GUPTA appears to be a low-to-mid level candidate, with 3 years of experience.
View more
Work Experience
Machine Learning Engineer II
Rapid Acceleration Partners
November 2021 - May 2023
- Developed Table Detection & Data Extraction system, achieving 80% accuracy rate for images/documents. Implemented Signature Extraction & Matching solution. Established ML infrastructure, improving development workflows by 30%. Development of Question Answering System, document data extraction & entity extraction using transformers-based language models like BERT, DeBERTa, RoBERTa, LayoutLM, GPT. Automated & Optimized ML model training and inference, reducing processing time by 65%. Developed 80% accurate OpenCV ML algorithm for document data extraction. Classified Mortgage Docs, Invoices, and Resumes via Document & Token Classification. Managed end-to-end data pipeline from collection to validation.
Deep Learning Engineer
Experio Tech Pvt. Ltd
February 2021 - July 2021
- Engineered real-time traffic surveillance solution for state government, encompassing vehicle classification, helmet detection, seatbelt detection, triple riding detection, vehicle counting, and heavy vehicle wheel counting, achieving over 90% accuracy across all models. Designed and implemented unique OCR for Indian number plates the evaluated up to 85%, alongside devising a comprehensive data collection strategy with effective labeling management.
Founder & CTO
Dyota Labs
June 2023 - Present
- Engaged with clients, understanding business requirements, and building products to improve operational efficiency. Training & Integration of traffic surveillance models like CRNN based OCR, vehicle detection / classification, different violation detection, wrong lane identification, vehicle tracking with 2D Lidar, 4D Radar & ANPR Camera system Integration of multiple cameras for Automatic Number Plate Recognition (ANPR) real-time detection on Nvidia GPUs Achieving a remarkable 60 FPS model processing for 4D radar without external GPU & 150 FPS per camera for ANPR systems. Enhanced the performance of existing models, surpassing a mAP of 0.9 and reduced detection time of existing AI pipeline from 4 sec to 1 sec without GPU Robust deployment pipeline on edge devices such as Raspberry Pi, Nvidia Jetson, Intel NUC, and Rock Pi. Built products for improving operational efficiency for quality assessment for BPO. Integrated LLMs like GPT-4 and Claude for call summarization and fine-tuned custom LLMs for multilingual call summarization.
Education
Sharda University
B.Tech.
January 2017 - January 2021
