Jun 2025 – Present · 1 yr

Mar 2023 – Jun 2025 · 2 yrs 3 mos

• 1) Developed and Deployed the Recommendation Engine and Semantic search engine using Generative AI , LLM models, Prompt Engineering with Elasticsearch Document Store or milvus indexing in 1AAA Team of Accenture. Build a semantic search engine for Structured documents, Microsite text and image/videos. Performance optimization of existing search functionality. Knowledge graph creation across all Accenture Microsites.
• Core Functionality:
• Content scrapping - Data Extraction from Star-dog. Retrieve the data from Star-dog. Web scraping or web data extraction is used for data scraping from a website. Web scraping of web pages fetches and extracts web pages. Fetching is downloading a page.
• Here for extracting the data from URLs, we are using the bs4 (Beautiful Soup) library for web scraping.
• Content extraction - Extract the data from different data types like PDF, PPTX, EXCEL and DOCX. Using different Python libraries.
• Pre-processing of data - Data Preprocessing is a technique that is used to convert the raw data into a clean data set. Text preprocessing is an essential step in natural language processing (NLP) that involves cleaning and transforming unstructured text data to prepare it for analysis. It includes tokenization, stemming, lemmatization, and stop-word removal.
• Data Summarization - Test summarization is a Natural Language Processing (NLP) task that summarizes the information in large texts for quicker consumption without losing vital information.
• Embedding - Using the BERT tokenizer, creating word embeddings with BERT begins by breaking down the input text into its individual words or parts.
• Fine-tuning, Improving Cosine similarity score - To improve the cosine similarity score we tried different distil-bert models from huggingface. For fine-tuning, we explore the different methods.
• API’s -Create the api for the search , Backup creation , For retrain , For get logs.

Mar 2022 – Mar 2023 · 1 yr

• 1) Involving in Building AI Enabled Semantic Search engine.
• 2) It has the facility of human-in-loop functionality for language experts to review and update machine transcription and translation.
• Core Functionality Offered:
• Transcription: Video Translation Platform tool takes any given video and transcribes it’s audio to text to produce a transcript of the entire video for the viewer’s reference.
• Subtitles: The tool also creates a subtitle file for the transcribed video which can be played through the video player.
• Glossary: Video Translation Platform provides the option for custom dictionary in the form of glossary file, translation service should use domain-specific terminology provided in custom glossary.
• Translation: Translation lets you dynamically translate between languages selected based on your content needs.
• Expert Review: This is a provision in which a human-in-loop is there, which can update the AI transcribed and translated content to ensure it is consistent and as per the required standards for your project by using our built in easy to use “Expert Review” functionality.