Data Scientist at Manaknightdigital

Toronto, Ontario, Canada

As a data scientist, my primary expertise lies in leveraging data to solve complex business problems. With statistical analysis, data visualization, machine learning, Deep Learning and API's. Here is my various field of works:

• Machine Learning | Deep Learning | Data Analysis & Visualization
• Experience with Statistical Tools | Written & Verbal Communication
• Scripting Language: Python | Database: MySQL
• ML Algorithms and Deep Learning Algorithms
• NLP: Text Normalization | BERT | Sentiment Analysis | Word Embedding
• Data Visualization & Analysis: PowerBI | Excel
• Web Framework: Flask || Libraries: Tensorflow | Scikit-learn | Keras | Seaborn
• Time Series Analysis
• Data Fetching from GPT and Analysis Data

Skills: Natural Language Processing (NLP) · Large Language Models (LLM) · MLOps · PyTorch · TensorFlow · Exploratory Data Analysis · Computer Vision · Data Visualization · GPT-3 · Artificial Intelligence (AI) · Data Science · Time Series Analysis · Deep Learning · Machine Learning · Python (Programming Language) · Finetuning LLM's

Projects:

• Atlas Copco Chatbot
  Primary Goal: to create a chatbot that will answer the question’s based on the requirements using the given data.
  Solution: Collect the data(excel file) which contains various product info(Model Name, Price, Capacity, Products Links). Then used the GPT(GPT- 4) model to answer the questions of what the user need. Also, tackle the token size in the GPT models.
  Library: openai GPT-4 model, Flask, openpyxl, pandas, numpy, re, nltk, sklearn, matplotlib, seaborn, tensorflow, keras, etc.
  Website: Atlas Copco
  
• Fraud Transaction Detection
  Primary Goal: given a dataset of past transactions and then from this data set have to build an ML model which can detect fraud transactions.
  Solution: perform EDA and Feature Extraction apply multiple ML algo(Xgboost,SVC,Logistic Regression) then applied hyperparameter optimization and deploy on the server for production. Got 90% accuracy using Xgboost.
  Library: scikit-learn,Flask etc.
  
• Michel AI Parabroker ChatBot
  Primary Goal: scrape data from different lenders then get answer out the data from user query
  Solution: used Retrival Augmentated Generation(RAG) based searching techniques to get the answers and for data stroing used Pinecone also used Cohere Reranking to refine the RAG search
  Library: PyPDF2,beautifulsoup,gunicorn,GPT-4,Pinecone,RAG,Cohere,Flask
  Website: Michel
  
• Football Analyst ChatBot
  Primary Goal: to make a Football Analyst which will analyze the pass data for a perticular Team and Player and Predict the future match
  Solution: scrap data from different football data sources and then used RAG,Pinecone and Cohere to get the response out of the huge ammount of data
  Library: BeautifulSoup,PyPDF2, RAG, Pinecone, Cohere
  Website: Kaizenwin
  
• Image Generation(Aieventbooth)
  Primary Goal: user will add some presets for a given image then it will generate the image
  Solution: used Stable Diffusion Image Generation, trained on custom image and then based on the used presets generates the image
  Library: transformers,pytorch,gunicorn
  Website: Aieventbooth
  
• Ai Energy
  Primary Goal: client's have some organization's privacy and policy, now a tender will came and we have look up the organization's data how much match each of the tender's and why matched
  Solution: to solve this problem first chuck the data into different sections for this finetuned different models like Llama-2-7B and 13B also Mistral-7B then used the cosine similarity to get the matched data and then used the GPT-4 to get the answer out of the data
  Library: transformers,pytorch,gunicorn
  Website: Aienergy