Parikshit Parihar

I am an AI and Data Engineer with over three years of combined professional and research experience specializing in video analytics for safety-critical applications. I curate large-scale video datasets, define annotation guidelines, and build semi-automatic pipelines to ensure high-quality ground truth.

I lead stress-testing of models against real-world edge cases—occlusions, shifting light, and moving shadows—to validate robustness for real-time deployments. I evaluate model performance using metrics like accuracy, precision, recall, confusion matrices, and FPR/FNR, and translate findings into actionable improvements.

In my current role I partnered with ML engineers and product stakeholders to deliver dataset-aligned PRDs, debug models, and drive a documented 52% accuracy uplift through targeted interventions. I have hands-on experience with annotation tools and reduced manual annotation effort by 70% via pipeline engineering.

My background includes research projects in socioeconomic analysis and large-scale algorithm comparisons, and I bring strong Python, SQL, cloud, and data analysis skills to roles that demand rigorous evaluation, operational robustness, and cross-functional collaboration.