• Over 6+ years of experience in AI, specializing in machine learning and deep learning models, as well as generative AI. Skilled in web frameworks like FastAPI and programming languages, including Python.
• Worked with a variety of databases, including SQL, NoSQL, and VectorDB.
• Extensive experience with AI frameworks such as TensorFlow, Keras, PyTorch, Langchain, and Llama-Index.
• Proficient in cloud technologies like AWS and Azure, utilizing multiple services.
• Experienced in using cloud services like GCP VM, AWS EC2, S3, Lambda Function, SageMaker, and RDS.
• Skilled in Docker deployment for AI models and applications.
• Worked with RAG, DAG, and agent-based workflows, including pipelines using VectorDB.
• Expertise in Lambda functions and ETL operations for data streaming from Core-IoT to DynamoDB.
• Proven experience leading AI projects focused on large language models (LLM) and generative AI.
• AI/ML developer with expertise in Computer Vision, Machine Learning, LLMs, and Gen AI model development and production deployment.
• Utilized machine learning techniques to develop and evaluate algorithms for enhancing performance, quality, data management, and accuracy.
• Strong experience in Computer Vision, including Image Processing, Object Detection, Face Recognition, Pose Estimation, Image Segmentation, 3D Image Processing, and 3D Image Classification.
• Well-versed in using the Langchain framework for custom model deployment, as well as GPT-based model integration with VectorDB (e.g., Pinecone, PGVector, Chroma).
• Experienced in handling real-time (time series) data.
• Proficient in creating APIs and deploying AI-based models on servers using the FastAPI framework.

Programming Language & Frameworks: Python

Machine Learning: Expertise in regression and classification algorithms, including Linear Regression, Logistic Regression, KNN, etc.

Deep Learning: Skilled in CNN, ANN, TensorFlow, Keras, YOLO, OpenCV, Deeplab, Langchain, LLM, and GPT models

Deployment: Experienced with Docker, Azure VM instances, AWS EC2 instances, and local servers

API Development: Proficient in Flask and FastAPI

Version Control: Git, GitHub

Project Management Tools: Visual Studio Code, Visual Studio, Anaconda, PyCharm

Database: MongoDB, Redis, MySQL, PostgreSQL, PGVector, Pinecone, Chroma, Qdrant

Cloud Technology:

• AWS: Experience with EC2, S3, SageMaker, Lambda, DynamoDB, Core IoT, Textract, VPC, RDS, IAM, and Bedrock
• Azure: Skilled in VM, AI Search Service, AI Cognitive Service, AI Model Service, and Web App Service
• GCP: Proficient with VM

KAPO:
Industry: IT
Start: July 2024
End: November 2024

Project Description:

• Developed a secure, closed-network AI chatbot system for a client’s internal use within a restricted environment, without internet access. The solution features two distinct interfaces: a document management interface for administrators and a conversational interface for end-users.
• The AI chatbot enables employees to query an extensive, up-to-date knowledge base in German, created from documents uploaded by administrators.

Roles & Responsibilities:

• Developed a generative AI-based chatbot using Ollama’s Llama 3.2 model, supporting German-language queries and responses, while adhering to the client’s closed-network requirements.
• Designed a dual-module application: an Admin module for managing knowledge sources through document uploads, and a User module for chatbot interaction within the knowledge base.
• Configured ChromaDB as the local data storage solution for embeddings, facilitating optimized and relevant data retrieval without external dependencies.
• Implemented a Retrieval-Augmented Generation (RAG) pipeline using Nomic embeddings to enhance the relevance and accuracy of chatbot responses based on uploaded documents.
• Deployed the application using Docker for a secure and scalable environment.

Technologies Used: LLM, RAG, Docker, ChromaDB, Gen AI, Python, TypeScript, Llama-Index

Lallan AI:
Industry: IT
Start: March 2023
End: October 2023

Project Description:

• Developed a custom Large Language Model (LLM) capable of interacting with personal data.
• Utilized the Langchain framework for connecting the LLM model with VectorDB.
• Worked with agents, memory, model, and chain components within Langchain.
• Implemented a dynamic system where users can engage with both personal data and global admin data, enhancing the chatbot's versatility.
• Utilized GPT models and a custom model for generative language outputs and embeddings.
• Integrated VectorDB to store embeddings of text documents for optimized retrieval.

Roles & Responsibilities:

• Led the project, focusing on workflow definition and solution architecture.
• Developed a custom LLM using the Langchain framework, enabling dynamic interactions with both user-specific and global data for a personalized experience.
• Integrated VectorDB for efficient text embedding storage and retrieval, optimizing performance and enabling personalized responses based on user data.
• Implemented Langchain’s agent, memory, and chain components to support context-aware conversations and smooth transitions between various data sources.

Technologies Used: Python, Langchain, PGVector, LLM, Llama 2 (Q), GPT, Hugging Face, PostgreSQL

AI DB Talk
Industry: Insurance
Start: October 2023
End: February 2024

Project Description:

• Enables users to interact with SQL databases using plain language questions, eliminating the need for SQL expertise.
• Utilizes Azure OpenAI Service to interpret user intent and generate precise SQL queries from natural language inputs.
• Implements LlamaIndex to manage and streamline query generation and execution for efficient processing and quick response.
• Integrated seamlessly within Microsoft Teams, providing users with easy access to the chatbot without needing to switch applications.
• Capable of executing SQL queries, delivering results, generating visualizations like plots and charts, and handling follow-up questions to improve user engagement and comprehension.

Roles & Responsibilities:

• Designed and developed the chatbot's architecture, ensuring smooth integration with Azure OpenAI Service, LlamaIndex, and Microsoft Teams.
• Focused on crafting a user-friendly interface with advanced natural language processing capabilities, dynamic result displays, and visualizations (plots and charts) for effective data presentation.
• Conducted extensive testing to ensure accurate and efficient performance, resolving post-deployment issues, and continuously enhancing the system based on user feedback and changing requirements.

Technologies Used: Azure Web App Service, Azure OpenAI Service, Azure Bot Service, LlamaIndex, MS-SQL, Python, Plotly Express

Ask Liberty:
Industry: Fin-tech
Start: February 2023
End: August 2023

Project Description:

• Created a Gen AI-powered chatbot for employees, embedded within a private client website to facilitate intelligent information access.
• Leveraged Azure services to store and manage a comprehensive knowledge base of HR policies, regulations, and company-specific modules, accessible through user interactions.
• Implemented a system that automatically scrapes and updates the knowledge base by extracting content from specified modules on the client’s website at scheduled intervals using a cron job, ensuring the chatbot delivers the most up-to-date information.

Roles & Responsibilities:

• Designed and executed an automated data extraction pipeline using a web scraper, set to run on a scheduled Azure-based cron job to collect specified content from the client’s private website.
• Configured Azure Blob Storage to securely store extracted data, providing a scalable and centralized repository.
• Developed and deployed an Azure Function to process the stored data, triggering an AI-driven chatbot embedded on the client's private site.
• Integrated a Retrieval-Augmented Generation (RAG) pipeline within the chatbot, allowing it to respond to employee inquiries with information based on the most current data in Blob Storage.
• Optimized prompts for the LLM to ensure accurate and contextually relevant interactions, enhancing the overall employee experience.
• Managed the deployment process on Azure, utilizing CI/CD pipelines for efficient updates, testing, and integration, guaranteeing stable performance and ease of maintenance.

Technologies Used: Azure Web App, Azure Function, Azure Blob Storage, OpenAI (ADA002), OpenAI (GPT-4o), ChromaDB, Langchain

Face Recognition System:
Industry: IT
Start: October 2023
End: January 2023

Project Description:

• Developed a desktop application for attendance management using PyQt. Contributed to enhancing face detection and spoof detection components by leveraging the DLib library.

Roles & Responsibilities:

• Designed the pipeline for face detection and spoof detection.
• Designed the database schema and feature vector store for storing and retrieving facial data.
• Optimized the system and conducted performance measurement for better efficiency.
• Created documentation and reporting to support project maintenance and further development.

Technologies Used: Python, DLib, OpenCV, Postgres, TensorFlow

SafeScan – Face Mask Detection:
Industry: IT
Start: October 2023
End: January 2023

Project Description:

• Developed both a kiosk and desktop application to detect face masks, body temperature, and facial attributes (such as age, gender, and emotion) using the CyberLink SDK. Contributed to building the attendance system with PyQt for the graphical user interface (GUI).

Roles & Responsibilities:

• Designed and integrated a wrapper pipeline with the CyberLink SDK for face mask and body temperature detection.
• Developed the database schema and feature vector store for managing detected data.
• Optimized system performance for efficient operation.
• Authored documentation and provided training for system usage.

Technologies Used: Python, TensorFlow, SQLite3, Postgres, OpenCV, PyQt, PyInstaller, Inno Setup

Hyper Spectral Imaging:
Industry: Agriculture
Start: October 2022
End: January 2023

Project Description:

• This project focused on segmenting wheat plants, identifying regional properties, classifying plant stages, and counting wheat spikes using YOLO-v3. The second part of the project involved generating reflectance graphs from hyperspectral imaging captured by a hyperspectral camera.

Roles & Responsibilities:

• Performed wheat spike detection using YOLO-v3 for object detection.
• Applied Gradient Boosting algorithm for plant stage classification.
• Prepared the wheat spikes dataset for custom object detection.
• Conducted plant segmentation using image processing techniques with OpenCV.

Technologies Used: Python, Darknet, Visual Studio, CMake, OpenCV, PyQt, LabelImg, Gradient Boosting, KNN, Random Forest

Project: Gaze Tracker
Industry: Robotics
Start: May 2022
End: August 2022

Project Description:

• Developed a system for toy detection using object detection algorithms and gaze tracking with image processing. The project also included an alert system to notify when the toy was detected.

Roles & Responsibilities:

• Used YOLO V5 for object detection to detect balls.
• Integrated DeepSort tracker for tracking detected objects.
• Developed a custom model for eye gaze tracking.

Technologies Used: Python, Darknet, OpenCV, Playsound

Augmented Startup:
Industry: IT
Start: January 2022
End: March 2022

Project Description:

• This project focused on leveraging YOLO V4 for object detection, object tracking, and GUI applications. The following applications were developed as part of this project:

Roles & Responsibilities:

• Performed object detection using the COCO dataset.
• Implemented object (person) tracking using DeepSORT.
• Developed a social distance web application.
• Created an empty parking slot detection and car count application.
• Developed a mask detection application.
• Created both Windows and Linux applications for each of the above functionalities using PyQt5.

Technologies Used: Python, Darknet, Visual Studio, CMake, OpenCV, PyQt, LabelImg

3D Model Generation:
Industry: Art & Museum
Start: January 2022
End: March 2022

Project Description:

• This project aimed to provide an API service for generating high-quality 3D models of real-world objects, scenes, or artworks using Neural Radiance Fields (NeRF) technology. The API was designed to enable artists, museums, curators, and digital archivists to transform 2D images, videos, or 3D scans into photorealistic 3D models for various applications, including virtual galleries, immersive exhibits, online archives, and digital preservation.

Roles & Responsibilities:

• Developed the NeRF-powered 3D model generation pipeline.
• Integrated the solution with FastAPI.
• Designed and deployed a scalable and customizable solution.

Technologies Used: Python, Colmap, NeRF, PyTorch, Trimesh, Blender

Diabetic Eyes Detection:
Industry: Healthcare
Start: August 2021
End: February 2022

Project Description:

• The solution involves the development and deployment of deep learning models to automatically detect and classify diabetic retinopathy (DR) in retinal images. These models analyze eye scans such as fundus photographs, OCT (Optical Coherence Tomography) scans, and fluorescein angiography images to identify early signs of diabetic eye disease. The solution is designed to detect five stages of diabetic eye conditions.

Roles & Responsibilities:

• Developed models including ViT (Vision Transformer), custom CNN models, and U2NET for image segmentation.
• Managed ML flow pipeline and implemented MLOps for continuous integration and delivery.
• Utilized Python and FastAPI for model deployment.
• Deployed solutions on Jetson boards for edge processing.

Technologies Used: Python, FastAPI, Jetson Boards, ViT, Custom CNN Model, U2NET, ML Flow, MLOps

Revelio:
Industry: IT
Start: May 2021
End: January 2021

Project Description:

• The project focuses on providing solutions for various computer vision tasks, such as object detection, face detection, photo restoration, object segmentation, photo sketching, OCR (Optical Character Recognition), and photo carbonifying. Revelio offers both a web interface and an API service for tag and object detection. Using REST APIs, users can integrate the service into their projects and environments. Additionally, Revelio provides a management console for customers to review their plans, usage, and case details.

Roles & Responsibilities:

• Designed and developed machine learning models for various computer vision tasks.
• Integrated custom models into the platform for enhanced functionality.
• Utilized Django Rest Framework for backend development and API service creation.
• Implemented MongoDB (NoSQL) for data storage and management.
• Employed technologies like TensorFlow, PyTorch, and Detectron2 for model development and deployment.

Technologies Used: Python, Django Rest Framework, MongoDB (NoSQL), JavaScript, Pytorch, Detectron2, Azure API Management Service, TensorFlow

Golf Trolley:
Industry: Sports
Start: July 2021
End: November 2021

Project Description:

• The project involves the detection of different areas on a golf course. Satellite information of a specific golf course or hole is sent to the backend, and our task is to extract an image from Google Maps. On the extracted image, AI-based segmented inference is performed to identify playing areas and various obstacles on the golf ground.

Roles & Responsibilities:

• Performed semantic segmentation to detect entities on the golf course using a DeeplabV3 and ResNet50-based model.
• Utilized Azure server-based virtual machines (VM) and Azure Blob Storage for image storage.
• Developed a Flask-based API to receive latitude and longitude information for a specific golf course. Based on the received data, Google Maps API is called to extract the required image, and a deep learning model is executed on the Azure VM to return an overlay and segmented image from the model's results.

Technologies Used: Python, DeeplabV3+, OpenCV, Flask, Azure VM, Azure Blob Storage