Aditya Gaharwar

I’m an AI Engineer focused on building production AI systems using LLMs, computer vision, and backend automation, especially in noisy, real-world environments. I enjoy turning messy signals into dependable workflows—inspection automation, operational analytics, and multimodal data pipelines.

In my internship at Radials International Mining Services, I developed an automated attendance verification and payroll automation system using facial recognition and image metadata validation, reducing manual HR auditing effort by 60%. I also enhanced the pipeline with EXIF-based fraud detection, parallel processing (ThreadPoolExecutor), and local caching—cutting processing time from hours to minutes while improving authenticity checks.

To push AI reliability further, I built an automated image processing pipeline using Google Gemini’s multimodal API for a fixed 14-image truck inspection cycle, reducing manual audit workloads by 80–90%, and I architected a Tire Lifecycle Intelligence platform with FastAPI and Supabase RPCs to translate noisy telemetry into predictive maintenance directives, reducing downtime and asset breakdown costs by over 35%. Earlier, as a Research Intern at IIIT Dharwad, I co-developed a road segmentation pipeline from scratch (CVAT, a 1,000-image dataset), handled severe class imbalance with targeted augmentations and a hybrid loss, and tuned DenseNet-121 to reach 0.81 IoU and 98% validation accuracy.