Vishesh Ranpariya

developed a deep learning system to bypass text captchas using masked autoencoders+ vision transformers; developed a deep learning system to bypass text captchas using resnet45 + transformers; finetuned a yolo V-8 on custom dataset to bypass Google image captchas; made a semantic data fabric using Knowledge graph by designing the ontology with the help of an llm; made a Data catalog graph with a Knowledge graph; made a RAG pipeline with KG-routing of AI queries to its particular Database where it can find its data to search on; made Query planning agents with tool calling abilities and KG-routing abilities to find relevant Database

deployed a Web scraping project and data scraping project on a government tender website; deployed a knowledge graph rag automated QnA model for a specific database; made a QnA model utilizing locally running llm using ollama; finetuned a model using Qlora fine-tuning technique and made use of unsloth; used OCR to recognise tenders from documents and open-cv to extract the exact details; worked on VBA macros and proficient in excel; Made data pipelines from scraping the data using selenium to managing it and then storing it in databases and data warehouses; used OCR and open-cv(image-preprocessing) to bypass captcha on goverment website which helped in cost reduction of the company

Developed apps AI/ML and LLM apps using the langchain Framework; Developed the APIs for the following apps using Flask; Developed a custom data - AI chatbot using deep learning and reinforcement learning algorithms; worked on FAISS and Pinecone databases; used Gemini to build ai powered applications; made use of AI agents to make AI powered applications; worked in PyTorch ,TensorFlow, Keras ,OpenCV , mediapipe

Developed machine learning models and algorithms for predictive analytics. Constructed supervised and unsupervised learning models using Python libraries like Scikit-Learn, TensorFlow, and Keras. Utilized a variety of techniques including linear regression, logistic regression, decision tree, random forest, gradient boosting, and neural networks. Maintained software products including programs, webpages, and databases