• Over 6+ years of experience in Design and Development, and implementation of various client-server enterprise applications in Python using a web framework like Django, Django Rest Framework, Streamlit and hands-on experience in developing and deploying data analytics and complex AI/ML models in various domains.
• Proficient in various ML techniques such as regression, classification, clustering, Decision Trees and deep learning, Generative AI.
• Developed projects using Techniques like (RAG) Retrieval Augmented Generation, fetch the required data from vector DB.
• Hands-on experience on GenAI, LLM (Anthropic Claude, OpenAI, OLLaMA), Langchain, OpenAI API(chatgpt-4, fine-tune), NLP, GooglePaLM, LLaMA.
• Developed a RAG project using LangChain and Llama Index. (Indexing for vector db, similarity search in langchain).
• Good knowledge of converting PDF data to structured text into chunks and storing it in vector format into vector DB.
• Worked on Ollama using LangChain and Langserve and created APIs in FastAPI Framework.
• Good experience in Pandas, numpy, matplotlib libraries for data cleaning and pre-processing.
• Fine-Tuned pre-trained model for better accuracy and efficiency such as HuggingFace, BERT model.
• Have a knowledge of evaluation of RAG and diff. AI Model and how to optimize then using fine-tuning or improving RAG architecture.
• Have worked on some projects which involved NLP techniques (NLTK library) such as Tokenization, Part-of-speech (POS) tagging, Sentiment Analysis, Text Classification, Text-to-Speech.
• Good knowledge of OOPs (Object Oriented Programming) and applying Object-Oriented principles in the Software Development Life Cycle.
• Well versed with the design and development of the presentation layer for web applications using technologies like HTML5, CSS3, and JavaScript.
• Excellent knowledge in working with Django ORM and advanced concepts.
• Possess good communication, interpersonal and analytical skills, and a highly motivated team player with the ability to work independently and Team environments.

• Thinking of the different Machine Learning approaches and techniques according to the Business domain.
• Have a good command over data cleaning and preprocessing techniques on large datasets.
• Build Efficient Machine learning models. Can fine-tune the existing model for the needs and requirements and achieve good results.
• Gathering requirements and translating the business details into Technical design.
• Development of Python APIs for Mobile Apps and JS based Frameworks like ReactJS, AngularJS, etc.
• Designed and developed the UI of the website using HTML, CSS, and JavaScript
• Implemented REST API's in Python using frameworks like Django.
• Involved in Developing RESTful & micro-services using Python frameworks.
• Wrote and executed various MySQL database queries from Python using Python-MySQL connector.
• Followed the Agile methodology to develop the application.
• Using the GIT version control tool to coordinate team-development.
• Maintained and updated the application following the client requirement.
• Involved in organizing meetings to know the needs of clients for the Enterprise solution Implementation.

Languages/Scripting: Python, JavaScript, HTML5, CSS3
Python Frameworks: Django, Django REST Framework, StreamLit
AI: RAG (Retrieval Augmented Generation), GenAI (pipelines), ML, DL, NLP, LLMs(OLLaMA, Anthropic Claude), OPENAI (GPT 3.5 & 4), GooglePaLM, LLaMA, Whisper AI, MLN, Pandas, Numpy, Tensorflow, Pytorch, scikit-learn, cloud server
Python Libraries: Celery, django channels, etc
Web Services: REST API, OAuth & OpenID Connect
Databases: MySQL, PostgreSQL, MongoDB
Tools: Git, Docker, Jira, Trello, Asana
Cloud Server: AWS, Digital Ocean, GCP, Azure
Operating Systems: Linux Variants, Windows Variants
Technical Skills: Python, AI/ML, LLMs, Django, Django REST Framework, Fast API, Server level configurations, Data Structures, HTML, CSS, Javascript, jQuery.
Organizational skills: Leadership Qualities, Management, Entrepreneurship, Team work.

OKBase: Document matching (Lead Developer):
Tech Stack: Python, Langchain,LLMs(Opus, Haiku, Sonnet-3.5, gpt-4o),Graph DataBase(Neo4j), Cosine Similarity, fastapi,Pandas,llama-index,etc
Description:

• Converted PDF documents into structured JSON data for analysis and storage.
• Utilized Neo4j to store and manage complex relationships between document entities.
• Developed efficient methods to fetch and query data from the Neo4j graphdatabase.
• Implemented document similarity matching using LLM-generatedembeddings and cosine similarity.
• Created RESTful APIs with FastAPI to handle document processing, storage, and retrieval.
• Leveraged Pandas for data manipulation and analysis to support document processing and matching.
• Employed Llama-index to extract text from pdf for further processes.

Text to Image:

Tech Stack: Python, fastapi,streamlit openai(Dalle-3), Fine-Tune ,cloud service,etc
Description:

• Environment Setup: Install the required libraries including FastAPI, Streamlit, and OpenAI's API library. Ensure that you have access to DALL-E 3 through OpenAI's API.
• Create a FastAPI Backend: Set up a FastAPI application to handle image generation requests. This application will take in text input, call the DALL-E 3 API to generate images, and return the images as responses.
• Integrate with DALL-E 3: Write a function in FastAPI app that communicates with the OpenAI API, sending the text input to DALL-E 3 and receiving the generated image.
• Develop a Streamlit Frontend: Build a user-friendly interface using Streamlit, where users can input text and see the generated images. Streamlit will send the text input to the FastAPI backend and display the returned images.
• Deployment: Deploy the FastAPI and Streamlit applications on a cloud platform or server. Ensure that they are properly connected so that the frontend can interact seamlessly with the backend.
• Testing: Test the complete system by entering various text inputs and verifying that the images generated by DALL-E 3 match the descriptions.
• Optimization: If needed, fine-tune the interaction between text inputs,prompt and image outputs by adjusting parameters or refining the text descriptions sent to DALL-E 3.

Image to Image with prompt:
Tech Stack: Python, fastapi,streamlit openai(Dalle-2), Fine-Tune ,cloud service,etc
Description:

• Python: Core programming language for integrating various components and services.
• FastAPI: Utilized as the backend framework for handling API requests and managing the generation process.
• Streamlit: Frontend interface for user interaction, allowing users to input prompts and view generated images.
• Masking: Creating masking of image to change on specific part with prompt
• OpenAI (DALL-E 2): Image generation model used to transform the input image based on the provided prompt.
• Fine-Tune: Applied to enhance or specialize the model for specific image generation tasks.
• Cloud Services: Employed for scalable processing and storage, ensuring efficient handling of image data and model operations.

Stock Price Prediction and Decision-Making:

Tech Stack: Pre-processing the dataset and used ML models, LSTM, ARIMA, SARIMA and etc

Description:

• This project aims to develop a prediction model that forecasts future stock prices based on historical open prices and provides actionable investment decisions— whether to buy, sell, or hold stocks.
• Explore various algorithms such as Linear Regression, Decision Trees, Random Forests, Gradient Boosting Machines, and Recurrent Neural Networks (RNNs), particularly Long Short-Term Memory (LSTM) networks, for time series forecasting.
• Use evaluation metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and R-squared. Perform hyperparameter tuning and cross-validation to enhance model performance.
• Develop an application or API to provide users with predictions and actionable recommendations.

Roles & Responsibilities:

• Client communication (timely updates)
• Code quality and review and version controls on github
• Pre-Processing (ETL pipeline) of Data
• Split data into train & test dataset
• Implementing several ML/DL model depending on requirement and R&D (i.e. Linear Regression, ARIMA, Decision Tree, ANN, CNN, RNN and etc)

SQLify: Next-Gen Natural Language Query Engine (Lead Developer):
Tech Stack: Python, langchain, GooglePalm(LLM), VectorDataBase(Chroma), HuggingFace Embeddings, etc
Description:

• The project integrates cutting-edge LLMs, GooglePalm to comprehend natural language queries and generate SQL queries with human-like fluency and accuracy.
• These models are fine-tuned on large datasets of SQL queries and corresponding textual descriptions, ensuring robust performance across Electronic devices and tools domains.
• The system employs a sophisticated query generation pipeline that transforms natural language input into executable SQL queries.
• This pipeline involves LLM model, db-connection, sentence embedding, storing embedded vectors, syntactic parsing, semantic analysis, and custom query setup template,etc ensuring the coherence and correctness of generated queries.

Soccer video analysis (Lead Developer):
Tech Stack: Python,Django,Computer vision, deep learning yolo,html,css, javascript, GCP- gcs, vm instance
Description:

• Analysis and Tagging of players in video ,designed to enhance the understanding and interaction with other players.
• The system aims to analyze the behavior and actions of Player in video, extracting valuable insights to improve and teach to Players for their action.

Recommender system(Content-based):
Tech Stack: Python, Pandas, linear regression, Correlation,Vectorization ,Feature
Engineering , word embedding
Description:

• A fully functional personalized recommender system capable of generating accurate and relevant recommendations based on user preferences and behavior.
• Improved user engagement and satisfaction through personalized content discovery, leading to increased user retention and loyalty.
• Insights into the effectiveness of different recommendation algorithms and strategies in various contexts, informing future enhancements and iterations of the system.

Electronic and Fashion Recommendation to the user on E-commerce site (Lead Developer):

Tech Stack: Cosine similarity, F1 Score, collaborative filtering technique
Description:

• Developed a system which recommends electronic and fashion items to users based on their previous preferences.
• The recommendation system is built using cosine similarity and F1 score to ensure accuracy. The Amazon dataset is used for this purpose, and data preprocessing is carried out to clean and transform the data into a usable format.
• The data preprocessing step involves cleaning the Amazon dataset by removing irrelevant and duplicate data, handling missing values,and transforming the data into a format that can be used for modeling.
• The recommendation system employs a collaborative filtering technique that analyzes the user's past behavior and the behavior of similar users to generate recommendations.
• Specifically, the system computes the cosine similarity between the user and the items, and recommends items that are most similar to the ones the user has already liked.
• To evaluate the accuracy of the system, the F1 score is used as a performance metric.
• The F1 score combines precision and recall to measure the system's ability to correctly recommend items that the user is likely to be interested in.

Topic Modelling using Machine Learning:
Tech Stack: LDA, NMF, NLP, Stop Words
Description:

• To predict the categories based on the dataset provided.
• Generation of word clouds used the Latent Dirichlet Allocation (LDA) and Non-negative Matrix Factorization (NMF) Machine Learning model on the text to identify the most important topics or themes in the text data.
• To visualize the topics, word clouds were generated for each topic.
• Word clouds represented the most frequent words in a topic, with larger font size indicating higher frequency.
• Also used K-Means clustering ML algorithms for grouping of similar categories based on the prediction.