Kunal Singh Rajpurohit

Designed and deployed a fully autonomous AI Agent leveraging LangChain, Playwright, GPT-4, and Chrome DevTools Protocol (CDP) to scrape and understand project listings, initiate job applications, handle real-time recruiter interactions via smart prompt chaining, and complete assignments. Achieved 82% task automation rate and reduced manual involvement by 70 hours/month. Integrated Node.js microservices, Redis queues, and MongoDB for scalable workflow management. Used WebSockets and Event-Driven Architecture to maintain persistent communication pipelines with job platforms. Embedded feedback loops using Reinforcement Learning from Human Feedback (RLHF) to improve conversation handling and delivery accuracy. Orchestrated end-to-end CI/CD pipelines via GitHub Actions, Docker, and Kubernetes for production-grade deployment.

Designed and deployed a fully autonomous AI Agent leveraging LangChain, Playwright, GPT-4, and Chrome DevTools Protocol (CDP) to scrape and understand project listings, initiate job applications, handle real-time recruiter interactions via smart prompt chaining, and complete assignments. Achieved 82% task automation rate and reduced manual involvement by 70 hours/month. Integrated Node.js microservices, Redis queues, and MongoDB for scalable workflow management. Used WebSockets and Event-Driven Architecture to maintain persistent communication pipelines with job platforms. Embedded feedback loops using Reinforcement Learning from Human Feedback (RLHF) to improve conversation handling and delivery accuracy. Orchestrated end-to-end CI/CD pipelines via GitHub Actions, Docker, and Kubernetes for production-grade deployment.

Developed an end-to-end deep learning pipeline using Bidirectional LSTM networks and advanced time-series feature engineering for real-time elderly presence detection from smart meter data, achieving 96.45% accuracy in live simulations. Built and fine-tuned a Variational Autoencoder (VAE) for unsupervised anomaly detection, leveraging KL-divergence minimization to reduce reconstruction loss by 32%, improving system resilience against power signal noise and irregular patterns. Deployed containerized ML services on AWS (EC2) using Docker, orchestrated production workflows with GitHub Actions CI/CD, and integrated Grafana for real-time MLOps monitoring and performance degradation alerts.