Ramesh Baral

1) Designed and developed Cross-system Personalization and Recommendation solutions for Financial Portfolio contents and financial knowledge/learning contents. 2) Designed and developed Generative AI (GenAI) products using Large Language Models (LLM) for Financial and Marketing use cases. 3) Designed and developed propensity models for customer leads ranking.

1) Designed and developed Graph Neural Network (GNN) based Hybrid Recommendation System by integrating item content, metadata, and user-item interaction features. Used Pytorch, DGL, PySpark. 2) Designed Machine Learning, NLP, Deep Learning models for movie metadata (e.g., sub-genres) prediction from text data (description, plots, synopsis) and images. Integrated LDA, Word embeddings into deep neural networks and Ensembled models. Used Python, Scikit-learn, Spacy, Glove, VGG, Keras, Tensorflow. 3) Designed and developed a system to predict kids friendliness of movies using image and text data. Integrated text and image features into different state-of-art deep learning and transfer learning techniques. Used Python, Keras, TensorFlow. 4) Designed and developed preference elicitation questions generator using latent factor models. This was used to address the cold start problem in a movie recommendation system. Used Python, PySpark, Scikit-learn. 5) Designed music programs and radio stations recommender using different music metadata, broadcast metadata, and content-based features. Used Python, Spark, AWS EMR, S3. 6) Designed, developed a knowledge graph for movie metadata. Applied several data science algorithms, such as community detection, node centrality, link prediction, influencing nodes, ranking, node embedding, etc. to ensure data quality and discovery. Used Python, Neo4j, and Shell Scripts. 7) Prototyped Video recommender using AWS Personalize and Google Cloud Platform (GCP) Retail API

1) Hierarchical Point-of-Interest (POI) recommendation using geo-spatial, social, temporal, and categorical features of POI check-ins. 2) Explainable recommendation– Extracted aspects from >1M reviews, used Deep Learning to predict the sentiment polarity of review text, used factorization machine for recommendation. Used Pandas, Numpy, Convolution Neural Network, Stanford Core NLP, Word2Vec embeddings. 3) Multi-aspect personalized recommender system – Scalable recommender system based on matrix factorization and personalized ranking algorithm to rank and recommend locations to users. Used Python, Numpy, Pandas, Networkx, Apache Hadoop. 4) Exploiting roles of aspects for Personalized POI Recommendation - An analysis of categorical, social, spatial, and temporal aspects for personalized POI recommendation. Designed different recommendation models using Personalized Page Rank and Matrix Factorization and evaluated the performance of recommendation on different aspects (social, categorical, temporal, and spatial). 5) Spoken Dialog System for Humanoid Robots - Used Markov Model to design the spoken dialog system and integrated it with the Nao (Nao (pronounced now) is an autonomous, programmable humanoid robot). 6) Guest lecture for a graduate course "Introduction to Data Science". 7) Lab Instructor: Programming in Java I and Java II, Computing for Business course for undergraduate students.

1) Part time freelancer at UpWork (former Odesk) to design and develop machine learning and deep learning models, such as Gated Adversarial Networks (GANs) for anomaly detection, customer segmentation, and text categorization. 2) Designed and developed several enterprise software modules using Java Swing, JSF, and RichFaces. - UpWork profile https://www.upwork.com/o/profiles/users/_~0147f623447e5265fd/

Summer intern for SIPARCS (Summer Internships in Parallel Computational Sciences) program at NCAR 1) Led a team of three members in the research, design, and development of a RESTful microservice based climate data analytics and hosted it on Amazon EC2 and S3 storage. Designed a generic JSON based workflow spec using Celery and Python based workflow spec adapter for Girder Worker engine.

1) Saved ~50% of effort (~20hrs/week of ~20+ people) by designing and developing a visual dashboard to track and manage the file transfer process of GlobalScape EFT 6.3. 2) Visited head office (VersikHealth Inc, Waltham USA) from Oct 2012 - Feb 2013 to lead the design and enhance the file acquisition system, and gained ~35% (~5 mins/GB) speed in file acquisition and management by customizing Advanced Workflow Engine (AWE) of GlobalScape EFT 6.3 and using Powershell 2.0. 3) Saved ~50% of the human effort (~20hrs/week of ~100+ people) over the traditional file acquisition system by the design and development of automated data acquisition. Used GlobalScape EFT 6.3, Advanced Workflow Engine (AWE), and Powershell 2.0 scripting.

1) Saved ~30% (~1 min/ticket) of tickets adding effort by designing and developing an email interface for tickets addition. Developed a Issue Tracking system (with web and email interface) which was one of the heavily used in-house tools. Used JSP, Servlets, Oracle 10g, and WebLogic server. 2) Achieved ~20% (saved ~5 mins/GB) performance in the file download process by designing a multithreaded model. Designed and developed a Data Management and a File Acquisition system that used GlobalScape's FTP API for the client end (file acquisition) and JSP, Servlets, WebLogic server for the server end (web interface). 3) Saved ~12.5% of human effort (5 hrs/man/week) by developing an enterprise Project Management system to keep track of the status of data engineering projects. Used J2EE Spring MVC, and Oracle10g. 4) Saved ~70% of human effort (1hr/man of 400+ people) spent in employee appraisals, via the design and development of an Employee Appraisal System. Used J2EE Spring 3.0, Rich Faces, Hibernate ORM, and Oracle 10g. 5) Designed and developed a Project tracking system which was used to track the events related to a project life cycle. Used J2EE Spring Web Flow 2.0, Rich Faces, and Oracle 10g. 6) Developed and maintained the BankServ's GFXW (Global Fund Exchange Web) system – a web-based modern client/server application to facilitate the automation of the secured wire transfer in real time communication to the central bank’s wire operation department. Used J2EE Struts 1.0, Oracle 10g, and JBOSS server.

1) Developed Online Payment Gateway system to facilitate utility bill payments. Technology used J2EE Struts 1.0, Oracle 9.0. 2) Designed and developed several other J2EE based in-house tools.