Rohit Mehra Email & Phone Number

Seattle, Wa, Us

Artificial General Intelligence - Core Edge Infrastructure- MLOps for Edge Models.Automatic Speech Recognition - Core Transformers Group.- Built distributed model hosting service with SageMaker, ECS Fargate and DynamoDB.- Designed and built CI/CD system for model and inference engine deployments.

Framingham, Ma, Us

Contributions• Developed a BERT based Named-Entity Recognition tool to extract product attributes from title and description. Extracted attributes are used for category analytics, site content improvement, assortment planning, and product matching.• Cross-Modal Retrieval in the E-commerce context: Learning Semantic Text-Image Embeddings (MLE - Python, TensorFlow, scikit-learn, Azure)• Distributed Warehouse Management System (Full-Stack - Java, Android, Python, Django, SQL)Primary Responsibilities• Developing applications in the areas of Machine Learning, Order Fulfillment, Assortment Optimization, Warehouse Management, and Competitive Analysis.• Work closely with key business stakeholders to derive application requirements, set expectations, and communicate progress.• Hands-on ownership architecting and developing highly scalable & highly available applications in the Cloud.• Participate in the entire SaaS life-cycle driving adoption and scale applying best practices that drive agility and quality.

San Francisco, Ca, Us

Developed a stream processing using kafka-python to track people (user input images of target) in the wild over multiple video streams.

San Antonio, Texas, Us

• Worked with Advanced Analytics and Intelligence (A2I) team on the electrical outage prediction project.• Implemented ETL procedures in PySpark to integrate raw weather data into Hadoop Cluster.• Developed end-to-end distributed deep learning pipeline using BigDL framework for electrical outage prediction.• Built a Fully Convolutional Network (FCN) to predict electric outage frequency and durations geographically mapped over San Antonio using weather images.

San Antonio, Tx, Us

Key responsibilities are to implement, evaluate and document deep learning model architectures from research papers.• Implemented CapsNet (Python, TensoFlow) with dynamic routing mechanism from the research paper. Research in using Capsule routings to solve visual perception-based problems such as object localization, detection, and segmentation in a single-shot, sample-efficient, end-to-end deep learning pipeline.• Experimented with Stance Detection using Bi-directional Conditional encoding for Fake News Detection in microblogs. In-depth data analysis and feature engineering to understand language constructs and sentiment flow in fake news stories.• Designed an ensemble model to tackle duplicate question pairs challenge on Kaggle. The model comprised of 110 hand-crafted features for XGBoost and a Siamese network variant using Stanford GloVe embedding. The model was ranked among the top 4% in private leader-board with the log loss of 0.14 on the testing set.• Remodeled UNet architecture for object-background segmentation task. It was trained on Carvana Car Dataset, available at Kaggle. Convolutional Denoising Autoencoder (CDAE) was trained to remove noise from the predictions. This combination scored 0.9967 on the mean dice coefficient evaluation metric.• Investigated persuasion techniques used in natural language, with an aim to use them as features for the detection of phishing emails. For evaluation, developed a voting ensemble classification model which uses Gain-framed and Loss-framed persuasion techniques as features. These Identified features significantly outperformed other work done in this field, with the F-Score of 0.87.• Developed crawlers using Scrapy for e-commerce websites to collect prices, reviews and ratings for price comparisons and price forecasting using machine learning. These crawlers were deployed on servers and collected data for over 3 months.

Delhi, New Delhi, In

Developed USB Joystick to Serial Communication GUI which can be used to control the motion of a Gimbal. It was developed using VC++.

Houston, Texas, Us

Developed an Online banking System Using J2EE Technology.

India, In

Worked on LINUX and learned network management skills such as Telnet configuration, User Administration, File Systems, Management Disk quota & Partitioning, Routing, FTP configuration, DNS configuration, DHCP configuration, Samba configuration, Apache web server, Configuring Send mail, Firewall Services and Shell Programming.

Master’S Degree, Computer Science

Nanodegree, Artificial Intelligence

Bachelor’S Degree, Computer Engineering

High School, High School/Secondary Diplomas And Certificates