Ankit Tejwan

Results-driven Deep Learning Engineer with 2+ years of industry experience building and deploying production-grade Computer Vision and AI systems. Proficient in PyTorch, TensorFlow, YOLO-family models, and NVIDIA GPU acceleration (CUDA, TensorRT, cuDNN). Hands-on experience with real-time inference pipelines, FastAPI-based ML services, and edge/cloud deployment.

Strong foundation in machine learning fundamentals, model optimization, and applied NLP/LLM experimentation. Seeking to leverage expertise in ML, Computer Vision, and scalable AI systems to contribute to Amazon's Speech and Language AI team.

Computer Vision Application Developer - Allround (India) Vegetable Processing Machines Pvt Ltd - Ambala, Haryana

(2025-03)

• Developing a real-time potato quality inspection and defect classification system achieving measurement accuracy of ±2–3 mm and high-speed inference under 100 ms using NVIDIA CUDA, TensorRT, and cuDNN.
• Designing end-to-end ML pipelines integrating industrial cameras and YOLO-based object detection for automated defect grading at scale.
• Optimizing deep learning models for deployment on NVIDIA RTX 30/40/50 series GPUs, ensuring reliable production throughput.

Deep Learning Engineer - ADD Innovations Pvt Ltd - Greater Noida, UP

(2024-01 - 2025-02)

• Built and maintained production computer vision systems for manufacturing quality control, integrating YOLOv5 and YOLOv8 PyTorch models with NVIDIA CUDA and ONNX Runtime for both GPU and non-GPU deployments.
• Collected, annotated, and curated large-scale datasets using industrial cameras (USB, GigE — Basler, Hikrobot, Baumer, FLIR) and annotation platforms (Roboflow, LabelImg, MakeSense.ai).
• Deployed FastAPI-based REST APIs for real-time inference, enabling seamless integration with client manufacturing systems.
• Collaborated with cross-functional teams to align model architecture choices with client latency, accuracy, and hardware requirements.

Research Intern – Artificial Intelligence - Center for Advanced Studies, AKTU Lucknow - Lucknow, India

(2022-06 - 2022-08)

Supervisor: Prof. M. K. Dutta

• Designed and trained a custom CNN model for plant disease classification using TensorFlow on datasets from Kaggle and Papers with Code.
• Applied advanced model optimization techniques including data augmentation, batch normalization, and dropout regularization.
• Conducted systematic hyperparameter tuning experiments with Adam, SGD, and RMSprop optimizers to improve training convergence and generalization.

B.Tech - Information Technology - Rajkiya Engineering College Bijnor (2019-08 - 2023-06)