Kunal Vaghela

Technologies: Python, TensorFlow, PyTorch, Hugging Face, GPT/BERT LLMs, React, Next.js, TypeScript, Node.js, MongoDB, AWS Lambda, SageMaker, Docker, Kubernetes, Cloudflare Workers/KV, REST APIs, Antd, Passport.js • Designed and implemented end-to-end AI/ML pipelines processing 10M+ data points/day, using TensorFlow/PyTorch for model optimization, increasing accuracy by 35% and reducing inference latency 40%. • Built production LLM systems using Hugging Face + GPT/BERT, deployed via SageMaker, serving 500K+ monthly requests with 92% user-approval score. • Engineered multilingual NLP models (15+ languages) using transformers, tokenizers, and fine-tuned BERT, reaching 89% F1 and powering global sentiment/entity extraction workflows. • Developed defect-detection CV models using CNN architectures + OpenCV, reducing manual QA errors by 28%, saving over $2M annually. • Built complete MLOps lifecycle using Docker, Kubernetes, CI/CD, drift monitoring, reducing deployment cycles from weeks to hours. • Led full-stack development of ReservationsCenter.com using Next.js/React and Node.js, architecting Priceline API integrations and high-performance TypeScript UI systems. • Implemented Affirm.js payment flows and secure booking payload generation for upstream Priceline endpoints. • Designed a multi-tenant Node.js aggregation layer normalizing inconsistent hotel vendor schemas, improving reliability and reducing reconciliation issues. • Integrated Cloudflare Workers + KV caching to reduce external API costs and cut response latency during peak loads. • Mentored 5+ engineers in ML architecture, PyTorch optimization, and full-stack best practices.

Architected end-to-end AI/ML pipelines processing 10M+ data points/day, improving accuracy by 35% and reducing latency by 40%. Built and deployed LLM-based systems (GPT/BERT via Hugging Face) on AWS SageMaker, serving 500K+ monthly users. Engineered multilingual NLP systems (15+ languages) achieving 89% F1 score for sentiment and entity extraction. Developed computer vision models (CNN + OpenCV) reducing QA errors by 28%, saving $2M+ annually. Established full MLOps lifecycle (Docker, Kubernetes, CI/CD, monitoring), reducing deployment time from weeks to hours. Led full-stack architecture for ReservationsCenter platform using Next.js, Node.js, and TypeScript. Designed multi-tenant aggregation systems for hotel APIs (e.g., Priceline), improving reliability and data consistency. Implemented Cloudflare Workers + KV caching, significantly reducing API costs and latency. Mentored 5+ engineers in ML systems, PyTorch optimization, and scalable architecture design.

Designed scalable data pipelines processing millions of records using Python and MongoDB. Developed and fine-tuned ML models using TensorFlow and PyTorch, improving model accuracy and efficiency. Built NLP solutions using transformer-based architectures (BERT), supporting multilingual datasets. Collaborated on full-stack systems using Node.js + React with TypeScript for production applications. Improved API performance and system reliability through optimization and caching strategies.

Built backend services using Node.js and REST APIs, handling data ingestion and basic business logic. Developed initial Python scripts for data preprocessing and automation workflows. Assisted in training basic ML models and learned TensorFlow fundamentals for internal use cases. Contributed to frontend components using React and improved UI responsiveness.