Data Science & Gen AI Intern at Orange

Developed an AI-powered chatbot using LlamaIndex and ReAct agent, integrating multi-agent workflows and prompt engineering. Designed robust RAG pipelines involving data parsing, embedding generation, and seamless indexing in a vector database to process complex content including text, tables, and images. Built a text-to-SQL agent using n8n and ReAct to query both CSV files and PostgreSQL databases from natural language inputs. Integrated and deployed a self-hosted LLM on on-prem servers using Kubernetes, ensuring scalable deployment and secure inference. Deployed the chatbot across internal platforms and Microsoft Teams using Docker, CI/CD workflows, Tock Studio, and on-prem VMs.

Student Job – Computer Vision & Machine Learning at ReaLift

Developed a hand tracking module for hand size estimation and keypoint extraction using Mediapipe and deep learning models. Created an interactive dashboard with Bokeh for real-time visualization and analysis of model outputs. Automated testing workflows and managed deployments using AWS Batch, Docker, and CI pipelines with AWS ECR.

Research Project – Artificial Intelligence and Data Science at IRIT

Built full data pipelines to collect and process individual activity data from the Fitbit API using PostgreSQL. Applied clustering (PCA, Hierarchical Clustering) to segment users and improve LLM-based recommendations using RAG. Modeled user stress levels with a custom PyTorch neural network, integrated into a user-facing app with Ionic, Django, and React.

Intern in ML and Computer Vision at ReaLift

Optimized a foot size prediction ML model by converting CPU-based OpenCV & NumPy operations to GPU using PyTorch. Improved masking methods using the Hugging Face ’Depth Anything’ model and performed extensive testing on AWS Sagemaker.

Data Science & Gen AI Intern at Orange

Developed an AI-powered chatbot using LlamaIndex and ReAct agent, integrating multi-agent workflows and prompt engineering. Designed robust RAG pipelines involving data parsing, embedding generation, and seamless indexing in a vector database to process complex content including text, tables, and images. Built a text-to-SQL agent using n8n and ReAct to query both CSV files and PostgreSQL databases from natural language inputs. Integrated and deployed a self-hosted LLM on on-prem servers using Kubernetes, ensuring scalable deployment and secure inference. Deployed the chatbot across internal platforms and Microsoft Teams using Docker, CI/CD workflows, Tock Studio, and on-prem VMs.

Student Job – Computer Vision & Machine Learning at ReaLift

Developed a hand tracking module for hand size estimation and keypoint extraction using Mediapipe and deep learning models. Created an interactive dashboard with Bokeh for real-time visualization and analysis of model outputs. Automated testing workflows and managed deployments using AWS Batch, Docker, and CI pipelines with AWS ECR.

Research Project – Artificial Intelligence and Data Science at IRIT

Built full data pipelines to collect and process individual activity data from the Fitbit API using PostgreSQL. Applied clustering (PCA, Hierarchical Clustering) to segment users and improve LLM-based recommendations using RAG. Modeled user stress levels with a custom PyTorch neural network, integrated into a user-facing app with Ionic, Django, and React.

Intern in ML and Computer Vision at ReaLift

Optimized a foot size prediction ML model by converting CPU-based OpenCV & NumPy operations to GPU using PyTorch. Improved masking methods using the Hugging Face 'Depth Anything' model and performed extensive testing on AWS Sagemaker.

Data Science & Gen AI Intern at Orange

Developed an AI-powered chatbot using LlamaIndex and ReAct agent, integrating multi-agent workflows and prompt engineering. Designed robust RAG pipelines involving data parsing, embedding generation, and seamless indexing in a vector database to process complex content including text, tables, and images. Built a text-to-SQL agent using n8n and ReAct to query both CSV files and PostgreSQL databases from natural language inputs. Integrated and deployed a self-hosted LLM on on-prem servers using Kubernetes, ensuring scalable deployment and secure inference. Deployed the chatbot across internal platforms and Microsoft Teams using Docker, CI/CD workflows, Tock Studio, and on-prem VMs.

Student Job – Computer Vision & Machine Learning at ReaLift

Developed a hand tracking module for hand size estimation and keypoint extraction using Mediapipe and deep learning models. Created an interactive dashboard with Bokeh for real-time visualization and analysis of model outputs. Automated testing workflows and managed deployments using AWS Batch, Docker, and CI pipelines with AWS ECR.

Research Project – Artificial Intelligence and Data Science at IRIT

Built full data pipelines to collect and process individual activity data from the Fitbit API using PostgreSQL. Applied clustering (PCA, Hierarchical Clustering) to segment users and improve LLM-based recommendations using RAG. Modeled user stress levels with a custom PyTorch neural network, integrated into a user-facing app with Ionic, Django, and React.