09/21/2026 updated
PS
100 % available
Lead Computer Vision & Machine Learning Engineer
Bengaluru, India
Only remote
MS Research in Electrical Engineering, IIT KanpurComputer Vision & Deep Learning
Expertise in building and fine-tuning state-of-the-art computer vision models including YOLOv8, SAM, ViT, and ResNet for video, image, and multimodal tasks. Development of end-to-end pipelines from data curation through model training, evaluation, and ONNX export for clinical, industrial, and wearable deployments.
Generative & Predictive Modeling
Application of generative modeling techniques such as VAE and MAML to augment low-data clinical pipelines, improving downstream model accuracy. Development of gaze estimation models achieving approximately 90% accuracy on large-scale in-house datasets.
ML Frameworks & Tools
Proficiency in PyTorch, Keras, MAML, VAE, Transformers, SAM, and Vision Language Models (SigLIP). Hands-on experience with OpenCV, Optuna, NVIDIA DGX, ONNX, AWS, Git (CI/CD), Model Compression (Quantization, Pruning, Knowledge Distillation), and Model Optimisation.
Video Understanding & Temporal Analysis
Development of video-based computational pipelines for temporal behavioural analysis, extracting scene-level understanding from sequential and multi-frame visual data.
Edge Optimisation & Deployment
End-to-end model lifecycle management including training, evaluation, quantization, and ONNX export for edge serving. Implementation of post-deployment validation to monitor latency and inference accuracy across mobile devices.
Data Engineering & Augmentation
Development of large-scale in-house datasets with novel augmentation strategies, boosting downstream model accuracy by approximately 6%. Collection and annotation of clinical video datasets with standardized schemas and data privacy protocols.
Programming
Programming skills in Python, MATLAB, and C (Basic).
Mathematics & Signal Processing
Knowledge in Linear Algebra, Signal and Image Processing Fundamentals, Probability, and Random Signals.
Research Infrastructure & Hyperparameter Optimisation
Automated hyperparameter optimization using Optuna on NVIDIA DGX, significantly improving model stability and reducing R&D tuning time.
Expertise in building and fine-tuning state-of-the-art computer vision models including YOLOv8, SAM, ViT, and ResNet for video, image, and multimodal tasks. Development of end-to-end pipelines from data curation through model training, evaluation, and ONNX export for clinical, industrial, and wearable deployments.
Generative & Predictive Modeling
Application of generative modeling techniques such as VAE and MAML to augment low-data clinical pipelines, improving downstream model accuracy. Development of gaze estimation models achieving approximately 90% accuracy on large-scale in-house datasets.
ML Frameworks & Tools
Proficiency in PyTorch, Keras, MAML, VAE, Transformers, SAM, and Vision Language Models (SigLIP). Hands-on experience with OpenCV, Optuna, NVIDIA DGX, ONNX, AWS, Git (CI/CD), Model Compression (Quantization, Pruning, Knowledge Distillation), and Model Optimisation.
Video Understanding & Temporal Analysis
Development of video-based computational pipelines for temporal behavioural analysis, extracting scene-level understanding from sequential and multi-frame visual data.
Edge Optimisation & Deployment
End-to-end model lifecycle management including training, evaluation, quantization, and ONNX export for edge serving. Implementation of post-deployment validation to monitor latency and inference accuracy across mobile devices.
Data Engineering & Augmentation
Development of large-scale in-house datasets with novel augmentation strategies, boosting downstream model accuracy by approximately 6%. Collection and annotation of clinical video datasets with standardized schemas and data privacy protocols.
Programming
Programming skills in Python, MATLAB, and C (Basic).
Mathematics & Signal Processing
Knowledge in Linear Algebra, Signal and Image Processing Fundamentals, Probability, and Random Signals.
Research Infrastructure & Hyperparameter Optimisation
Automated hyperparameter optimization using Optuna on NVIDIA DGX, significantly improving model stability and reducing R&D tuning time.
Project history
Leading AI-Automated Interpretation for Urine Dipstick Diagnostics at AIIMS Patna. Building a YOLOv8 (pad detection) + ResNet18 + LAB features (520-dim) pipeline for automated urine-dipstick diagnostics on smartphones. Owned end-to-end model lifecycle including training, evaluation, quantization, and ONNX export for edge serving. Worked with clinicians to translate ambiguous diagnostic requirements into clear technical goals.
Worked on Gaze Estimation, Gaze-Guided Depth Estimation Using Infrared Imaging, and Industrial Print-Defect Detection. Developed a VAE + MAML based gaze estimation model improving accuracy by approximately 12% over benchmarks. Designed a system for predicting distance of gazed objects using IR cameras on smart glasses. Built an end-to-end defect detection system integrating illumination setups with high speed cameras. Applied classical image processing (template matching, SIFT, SURF) for defect localization.
Worked on Early Detection of Autism Spectrum Disorder in Children and Lung Tumor Tracking. Developed a video-based pipeline using Pose Estimation to extract temporal behavioral features, achieving 95% diagnostic accuracy. Collected and annotated the first India-specific ASD video dataset (35 subjects). Reduced computational overhead by 30% through feature-level pruning. Developed real-time external surrogate visual tracking algorithms for radiation therapy guidance (GYTI-recognized work).