Kunal Kumar Singh

Deployed and maintained AI-driven applications onsite at government and defense locations including Sena Bhavan and DGIS. Developed secure NLP solutions including chatbots, voice assistants, and speech recognition tools for offline environments. Built and fine-tuned NLP models using Hugging Face Transformers, PyTorch, and TensorFlow. Pre-processed mission-critical data sets with tokenization, lemmatization, and POS tagging to build robust NLP pipelines.

Designed NLP classification pipelines, improving chatbot accuracy from 68% to 90%. Built computer vision surveillance (YOLO, RetinaNet, OpenVINO) deployed across 50+ cameras, reducing false positives by 15%. Fine-tuned LLMs (LLaMA, Mistral, Deepseek) for defense Q&A, improving accuracy by 20%. Led a 3-member ML team delivering cross-functional AI solutions for defense deployments.

Built predictive modeling pipelines (ARIMA, SARIMAX, Prophet, LSTM), improving forecast accuracy from 72% to 92%, yielding approximately $15M in savings. Automated preprocessing pipelines, reducing data processing time by 35%. Applied big data analytics (Spark, GCP, Hadoop) on 10M+ records to identify five key revenue drivers.