Vibha Vikram Rao

Developing an advanced AI-driven global supplier search platform.

• Investigating critical data points to enhance search efficiency, leveraging techniques such as Retrieval Augmented Generation with open-source language models.
• Expanding custom framework for data extraction and processing.
• Creating Airflow DAGs to streamline solution deployment.
• Conducting statistical evaluations for all models to monitor short and long term progress.
• Preparing both technical and non-technical presentations of implemented models for broader communication.

Machine Learning Technology Stack Development

• Responsible for building the Machine Learning technology stack from ground up including data ingestion, indexing, performing exploratory data analysis to understand the problem space better, building POCs, building pipelines for retrieval, model and pipeline evaluation.
• Building models and solutions for EU Taxonomy (Europe's flagship sustainable finance regulation for assessing if a company/investments are environmentally sustainable or not)

Building Cutting-edge Natural Language Processing (NLP) Systems

• Leveraged state-of-the-art large language models, including GPT3.5 Turbo and GPT4, in conjunction with cutting-edge frameworks like LangChain and Haystack, to engineer sophisticated Natural Language Processing (NLP) systems.
• Developed specialized domain-specific document chatbots capable of elucidating complex clauses within the taxonomy. These chatbots intelligently retrieve evidence from user-uploaded documents, providing detailed analyses against predefined criteria, while also generating comprehensive explanations.

Model Performance Evaluation and Enhancement

• Conducted meticulous evaluation of model and pipeline performance and continuously enhanced it based on user feedback.
• Built Evaluation system using Weights & Biases, and developed custom LLM assisted metrics to track model improvements.

Thought Leadership

• Demonstrated thought leadership as a speaker at PyCon 2023 in Berlin, delivering an enlightening presentation on harnessing the power of Haystack for climate-focused Question and Answer applications.

Machine Learning Model Development for Operational Excellence

• Engineered machine learning models across multiple business units to drive operational excellence. These models enabled the automation of critical processes such as search engines for customer support, tools to support associates to automatically fill forms, internal grammar correction tools for privacy.

Intelligent autonomous RFP Tool for Productivity Enhancement

• Pioneered the development of an intelligent tool capable of autonomously completing Request for Proposal (RFP) forms. This innovation resulted in a remarkable 40% reduction in associate workload, elevating productivity and increasing the potential win rate for bids.

Semantic Search Engine for Customer-Facing Team

• Enhanced the productivity of the customer-facing team by designing and implementing a semantic search engine. This engine harnessed advanced retrieval techniques, such as Dense Passage Retrieval, and incorporated Machine Comprehension to empower associates to quickly locate accurate information within internal company documents.

Grammar Correction Cognitive Tool Development

• Crafted a Grammar Correction Cognitive Tool by fine-tuning language models using parallel data sets containing grammatically correct and incorrect sentences. Employed data perturbation and back translation strategies to generate training data.

Object Detection Model Fine-Tuning for Data Extraction

• Fine-tuned an object detection model (YOLO) to extract regions of interest from forms, performed Optical Character Recognition (OCR) on these regions, and extracted key-value pairs from the forms. This experience demonstrates proficiency in computer vision and data extraction, which are valuable in AI applications.

Customer Feedback Analysis

• Conducted in-depth analysis of customer feedback by processing comments through a Sentence BERT model to obtain embeddings. Applied clustering algorithms to categorize comments, identifying trending issues for the IT support team.

- Built backend APIs for an employee referral portal

- Worked on building REST APIs to perform CRUD operations for an Airline Safety Product.