
Software Development Engineer-II
Truminds Software Systems · Sep 2023 – Present
- Architected a production AI pipeline integrating localization, OCR, masking, and defect detection for automated industrial quality inspection.
- Evaluated and benchmarked multiple transformer-based AI models including TrOCR, VisionEncoderDecoder, Swin+GPT2, ViT+TinyGPT, and Qwen3-VL to identify the most suitable solution for industrial OCR applications.
- Selected and deployed TrOCR for production after comparative evaluation, significantly improving OCR robustness under challenging imaging conditions.
- Built a modular AI inference pipeline combining transformer-based OCR and Computer Vision models for real-time industrial inspection.
- Optimized inference using ONNX, TensorRT, and INT8/FP16 quantization on the Qualcomm IQ9 Edge AI Platform, enabling low-latency production deployment with Azure ML pipelines.
- Developed intelligent document processing pipelines using FastAPI, PyTesseract, Selenium, Scrapy, and BeautifulSoup for automated extraction of structured and unstructured enterprise documents.
- Built scalable REST APIs supporting intelligent processing of PDF, DOCX, XLSX, and CSV documents while integrating MongoDB and Redis for efficient data management.
- Developed production AI solutions using YOLOv8 and SSD for automated retail compliance, product recognition, and cooler analytics.
- Managed the complete AI model development lifecycle for over 30,000 annotated images across 59 product classes, including dataset refinement, experimentation, model training, and performance evaluation.
- Improved detection accuracy to approximately 95% through augmentation strategies, model optimization, and iterative experimentation.
- Deployed scalable AI inference services using FastAPI, TensorRT, ONNX, and INT8/FP16 optimization for production environments.




