Headline

Senior AI Engineer @ Infilect | Computer Vision | Deep Learning

About

As a professional in the field of Deep Learning and Computer Vision, I have developed a strong understanding of the intricate details that drive these advanced technologies. Over the years, I have gained a wealth of practical experience in designing, implementing, and optimizing deep learning models and algorithms to solve complex problems in various industries. My expertise in Computer Vision has allowed me to develop a deep understanding of the underlying concepts and methodologies that drive these technologies. I am passionate about applying this knowledge to develop innovative solutions that can be leveraged to drive real-world impact. As a professional, I thrive on working in high-performance teams and am always looking for new and exciting opportunities to collaborate with other industry experts. With my passion for problem-solving, a keen eye for detail, and strong work ethic, I am confident in my ability to make a valuable contribution to any team I work with. I am excited to be part of a community that is working to push the boundaries of Deep Learning and Computer Vision and am always looking for new and exciting opportunities to collaborate with others in this space.

Location

West Bengal

AI stack

aistudio

deepmind

hugging face

mastra

Work experience

Senior AI Engineer at Infilect, Feb 2025 - Present · 1 yr 3 mos, Bangalore Urban, Karnataka, India, -Built an end-to-end ML pipeline for product image clustering using EfficientNetV2 + CLIP embeddings (LoRA fine-tuned) and hierarchical agglomerative clustering with pre-clustering, reducing compute time by 30% and memory usage by 70%. Achieved 95.8% precision and 97.9% naming accuracy, reducing manual categorization effort by 80%+. -Developed a Mask2Former-based instance segmentation pipeline for fine-grained retail pattern recognition, including data validation, augmentation, and COCO conversion. Achieved 66.2 mAP (0.50:0.95) with 84.6 AP50 and 70.9 AP75. -Built a scalable Flask-based embedding microservice integrating EfficientNetV2 + CLIP with Qdrant vector search, enabling RAG-driven real-time similarity search, metadata management, and clustering for retail product discovery. -Fine-tuned Qwen3-VL-Embedding-2B using QLoRA (8-bit) and SubCenterArcLoss, leveraging LLM-generated product descriptions to improve similarity search and clustering performance.

Senior Machine Learning Engineer at Fulltrack AI (by Maiden), Jul 2023 - Feb 2025 · 1 yr 8 mos, - Led the QA team, providing guidance on in-app annotations to improve accuracy and efficiency. - Handled Deployment using AWS Infrastructure.

Machine Learning Engineer at Fulltrack AI (by Maiden), Jan 2022 - Jul 2023 · 1 yr 7 mos, - Developed a multi-class classifier that would classify. batsman hand (left/right/none) and also wicket-keeper (present/absent). - Developed a shot classification model that would classify 21 different kinds of cricketing shots, also if the shot is lofted/grounded and if it is aggressive/defensive. Video classification was done using CNN(ResNet18) + LSTM. - Developed an Audio Classification model using BLSTM in order to detect it there was a hit sound in the video or not. - Developed a shot similarity algorithm, that generates a sorted list of shots based on similarity score. The algorithm takes a single video as input and generates a list of videos based on how similar the video is with respect to the input video. The similarity algorithm is calculated based on factors like Speed, Spin, Swing, Release Point, Pitch Point, and Beehive Points. - Built annotation tools in PyQt and OpenCV, models were built using Pytorch with video sessions fetched from AWS S3 bucket.

Education

Master of Computer Applications - MCA, Computer Science, Pondicherry University

Bachelor of Science - BSc, Computer Science, St. Xavier's College (Autonomous), Kolkata

Public links

github: https://github.com/fear-the-lord