SATHWIK User

I’m a results-driven Machine Learning Engineer with hands-on experience in computer vision, NLP/LLMs, and AI data operations. I build end-to-end ML solutions from research to deployment, including mobile-optimized models and intelligent chatbots.

In my current role as an AI Training Data Specialist (Process Executive) with NVIDIA via Randstad, I annotate and validate real-world driving videos to train computer vision models for autonomous vehicles. I created precise object classifications and free-space boundary labels for Enhanced Grid Mapping (EGM) systems, and I validate logical consistency of Chain-of-Thought outputs to improve answer quality scores by identifying inconsistencies.

I also served as a QAAuditor, reviewing peer-labeled datasets, maintaining >95% accuracy thresholds, and approving data batches before model training. I collaborate with NVIDIA’s automotive AI team to refine labeling standards and improve annotation workflow efficiency by 15%.

Across projects, I deliver measurable outcomes: a BERT-based resume NER model with 89% F1, and a two-tower embedding model that increased relevant recommendations by 25%. I developed PlantDoc with 95.3% top-1 accuracy and reduced mobile inference time from 340ms to 90ms using TensorFlow Lite quantization, and I built a sign-language action detection system using a temporal CNN+LSTM with MediaPipe pose estimation, deploying with Flask/TFLite for sub-120ms latency.