Hirak Jyoti Kashyap

I've delivered AI training data at Indica AI across speech, audio, image, and e-commerce workflows. My work has supported ASR/NLU, computer vision, and product-market mapping pipelines with high-quality structured datasets.

For speech and audio intelligence, I've completed 1,250+ annotations across 300+ hours of call recordings, including speaker diarization, phoneme tagging, and intent and sentiment labeling. I maintained 95%+ inter-annotator agreement and contributed to a 12% lift in transcription accuracy.

In computer vision, I've annotated 8,000+ images across 15+ object categories using bounding boxes, polygon segmentation, XML, and COCO JSON. My 96%+ labeling accuracy supported an approximately 10% improvement in object-detection mAP.

I've also tracked 5,000+ product listings across Flipkart, Meesho, and Alibaba using Python-based scrapers, reducing manual research time by an estimated 60–70%. I bring engineering discipline, careful QA, and a practical focus on making data more reliable for model training and business decisions.