Welcome to my GitHub profile.
I am a Machine Learning Engineer with over four years of experience in computer vision, natural language processing (NLP), and time-series analysis. Proficient in Python, PyTorch, TensorFlow, and Scikit-Learn, I have designed and trained deep learning models for a range of applications, including medical image analysis, where I implemented 3D and 2D segmentation as well as keypoint detection for enhanced diagnostic accuracy. I have developed and implemented document parsing solutions using OCR, enabling efficient extraction of structured data from various document types. My expertise extends to building LLM-powered chatbots leveraging Retrieval-Augmented Generation (RAG) to provide intelligent and context-aware responses, optimizing enterprise-level AI solutions.
- Python
- Scikit-Learn
- Tensorflow
- Pytorch
- Keras
- Tensorflow Serving
- TorchServe
- Nvidia Triton Inference Server
- HuggingFace
- Langchain
- LoRA
- QLoRA
- PEFT
- DeepSpeed
- OpenAI
- Ollama
- Docker
- Git
- DVC (Data Version Control)
- MLFlow
- Optuna
- Pandas
- Numpy
- Dask
- Kafka
- asyncio
- HTML
- CSS
- Javascript
- FastAPI
- Flask
- DRF (Django Rest Framework)
- SQL
- MYSQL
- SQLite
- PostgreSQL
- MongoDB
- BeautifulSoup
- Selenium
- AWS
- GCP
https://www.credential.net/384193e5-12b2-4142-9329-fa6d878e6537#gs.sf64l1
Implemented a video analytics solution using object detection, object segmentation, and object tracking to monitoring customers entering store, counting customer near specific locations, generate heatmap of customers visited locations, cashier presence at counter, time for which cashier was empty.
Implemented an Open Source Large Language Model (LLM) using LangChain with Retrieval-Augmented Generation (RAG) to assist users with queries about specific organizations.
I developed VoiceRAG, a voice-based chatbot using speech recognition, text-to-speech, and RAG (Retrieval-Augmented Generation) with Mistral LLM. Users can have natural, real-time conversations and ask questions about company data, making it ideal for virtual assistants, business intelligence, and customer support applications.
I developed an advanced Retrieval-Augmented Generation (RAG) system utilizing LangChain, Ollama 25B, and Weaviate as the vector database. The system is designed to efficiently answer user queries by retrieving relevant information from client-provided data and generating context-aware responses using a powerful LLM (Ollama 25B).
Implemented a HuggingFace and OpenAI model for Text-to-SQL, enabling users to query the database in English and displaying the results in knowledge graphs.
Implemented and custom-trained a 3D U-Net model for breast cancer tumor segmentation from MRI data.
Implemented and trained a deep learning model for the segmentation of medical 3D images (CT scans) using the U-Net architecture with an EfficientNet backbone.
Implemented and trained a custom deep learning model for keypoint detection in medical 3D images (CT scans).
Implemented and trained a deep learning model for text and field detection in various documents using the YOLO architecture, and developed a complete pipeline for document parsing and Optical Character Recognition (OCR).
Developed an application that analyzes classroom videos and identifies teacher actions such as sitting, talking, walking, writing on the board, presenting, and pointing to the board.
Classified questions as genuine or not using a dataset provided by Quora by training a deep learning model to determine if questions are sincere or insincere.
Implemented models to predict price change with least error possible, including price motion of direction prediction to reduce loss during trading.
Classified questions as genuine or not using a dataset provided by Quora. Developed a Django app that allows users to upload custom datasets and train various machine learning models, featuring dataset visualization, exploratory data analysis, dataset cleaning, model parameter customization, and model result visualization.
Created an API that takes input images along with a person's height in inches to return body measurements. Utilized a pre-trained pose estimation model to locate the joints of the human body in the images and calculated the pixel distances between different joints.
- Selenium
- Tensorflow
- Keras
