Looking for a job

Joydeep Medhi

@joydeepmedhi

Lead Data Scientist at Lowe's building store-wide tracking and generative-AI systems that delivered $100M+ in realized impact.

India

Message

At Lowe's, I lead computer vision and generative-AI programs across store intelligence, associate tooling, inventory optimization, and checkout automation. The work has delivered $100M+ in realized impact, with $1B+ projected at full scale.

I architected multi-camera tracking for store-wide customer-journey analytics and built learning pipelines that improved object-detection mAP by 15% while cutting manual annotation by 40%. I also shipped a real-time customer service alerting system that drove $1M+ in incremental pilot revenue.

At Mercedes-Benz Research & Development India, I brought the MBUX Interior Assistant in-cabin perception stack from research to production, including face recognition and authentication. My work there included efficient vision models for embedded automotive systems, a NeurIPS 2020 workshop paper, and five filed patents.

Experience

Work history, roles, and key accomplishments

Current

Lead Data Scientist

Current

Lowe's Companies Inc.

Jul 2022 - Present (4 years 3 months)

Leads computer-vision and generative-AI programs across store intelligence, associate tooling, inventory optimization, and checkout automation. Delivered $100M+ in realized impact with $1B+ projected at full scale.

Computer Vision Generative AI PyTorch TensorRT Nvidia DeepStream Multi Object Tracking Semi Supervised Learning Active Learning RAG VLM LLM Docker Kubernetes CI CD

Senior ML Research Engineer

Mercedes-Benz Research & Development India

Dec 2018 - Jun 2022 (3 years 6 months)

Owned the in-cabin perception stack for MBUX Interior Assistant from research to production, including face recognition and authentication. Designed efficient CNNs for human pose estimation, gesture recognition, and tracking on embedded automotive ECUs.

Computer Vision Deep Learning CNN Pose Estimation Face Recognition Metric Learning Knowledge Distillation Domain Adaptation PyTorch Torch INT8 Quantization FPGA Distributed Training

Deep Learning Research Engineer

Silversparro Technologies

Jul 2018 - Nov 2018 (4 months)

Built CCTV retail models for person detection, tracking, and age/gender classification, achieving 0.85 ROC-AUC for age and 88% accuracy for gender via transfer learning. Improved Faster-RCNN convergence and mAP with KMeans anchor-box initialization.

Deep Learning Computer Vision TensorFlow Object Detection Transfer Learning Faster RCNN K Means

Research Intern

CBRC, KAUST

May 2017 - Jul 2017 (2 months)

Developed deep neural models for causative disease-variant classification on imbalanced genome data with ontology features, achieving ROC-AUC 0.967 and AP 0.964. Applied statistical feature selection and fine-tuning on HPC GPU clusters.

Deep Learning Genomics Ontology Feature Selection HPC GPU Clusters

Education

Degrees, certifications, and relevant coursework

Indian Institute of Technology Delhi

B.Tech & M.Tech (MS) Dual Degree, Mathematics and Computing

2013 - 2018

Activities and societies: Head TA, Computer Programming (60+ students, 2017); coursework in Machine Learning, Data Mining, Linear Algebra, Probability & Stochastic Processes, Optimization, Image Processing.

Pursued a dual degree in Mathematics and Computing, completing a master's thesis on Calculus on Normed Vector Spaces and serving as Head TA for Computer Programming.