Praveen Roy

Project: AI Career Coach with Conversational Capabilities

Summary:
Developed an innovative AI-powered career coaching platform leveraging natural language processing (NLP), machine learning (ML), and conversational AI to facilitate seamless career transitions.

Key Contributions:

• Job Description Clustering: Aggregated and clustered job descriptions using sBERT, GloVe, Word2Vec, OpenAI Ada and GPT embeddings, enabling the identification of similar job roles.
• Career Transition Navigation: Designed and implemented methods utilizing inter-cluster distances to facilitate navigation between clusters, empowering users to explore career transition opportunities.
• Conversational Chatbot: Built an LLM-based chatbot using Retrieval-Augmented Generation (RAG) to provide personalized career guidance and support, integrating memory and out-of-context detection mechanisms.

Technical Stack:

• Programming Language: Python
• Cloud Computing: Azure ML, Azure Functions, Azure Kubernetes Service, Azure Blob Storage
• Databases: MongoDB, Cosmos DB
• NLP Frameworks: LangChain, Weaviate, Pinecone
• AI Utilities: OpenAI APIs, Hugging Face Models
• Other Frameworks: FastAPI, Docker, Git, Github
• Python Packages: NLTK, Matplotlib, Pandas, Numpy, Scikit-Learn, Seaborn, Pydantic, Uvicorn

Key Skills Demonstrated:

• Natural Language Processing (NLP)
• Machine Learning (ML)
• Conversational AI
• Cloud Computing
• Data Clustering
• Chatbot Development

Project: Real-Time Quality Control of Extrusion in 3D Bioprinters

Summary:
Developed and deployed a real-time quality control system for 3D bioprinting extrusion, leveraging computer vision and deep learning to ensure precise bio-ink segmentation, measurement, and analysis.

Key Contributions:

• Bio-Ink Segmentation: Utilized plane of focus to segment bio-inks, distinguishing between foreground and background regions using YOLO, Mask-RCNN, Vision Transformers, and SegDecNet architectures.
• Feature Extraction: Classified segments, determined angle to extruder, and identified major backbone contour.
• Measurement Strategies: Developed and tested two measurement approaches: sparse optical flow and geometry/trigonometry-based methods.
• Real-World Correlation: Correlated measurements to real-world dimensions and performed regression analysis for error correction.

Technical Stack:

• Programming Language: Python
• Cloud Computing: Azure ML, Azure VMs, Azure CLI
• Industrial Camera: Teledyne FLIR Camera with Spinnaker SDK
• Deep Learning Frameworks: TensorFlow, Keras, PyTorch
• Computer Vision Libraries: OpenCV
• Data Analysis: Matplotlib, Pandas, Numpy, Scikit-Learn, Seaborn
• Labelling Application: Labelme
• Other Frameworks: Docker, Git, Github

Key Skills Demonstrated:

• Computer Vision
• Deep Learning
• Real-Time Image Processing
• Quality Control
• 3D Bioprinting
• Python Programming
• Cloud Computing
• Data Analysis

Project 1: Coarse-Grained Modeling of Protein Complexes - Cartwheel Protein

Summary:
Developed a coarse-grained model of the cartwheel protein, found in centrioles, integrating multiple sources of amino acid contact pair information using the Integrative Modelling Platform (IMP).

Key Contributions:

• Protein Modeling: Constructed a detailed model of the cartwheel protein, leveraging various data sources for amino acid contact pairs.
• Optimization: Utilized IMP for model optimization, refining the protein structure.

Project 2: Deep Learning Architectures for Polypeptide Analysis

Summary:
Investigated deep learning architectures (CNN and NLP) to extract features from 3D protein structures and amino acid sequences, establishing relationships between structural and sequence features.

Key Contributions:

• 3D Feature Extraction: Explored CNN-based architectures (ResNet, VGG-Net, GoogLeNet) to extract features from 3D protein structures.
• Sequence Feature Extraction: Utilized NLP-based architectures (LSTM) to extract features from amino acid sequences.
• Structure-Sequence Relationship: Determined correlations between 3D structural features and sequence features.

Technical Stack:

• Programming Languages: Python, C++
• Deep Learning Frameworks: Keras, TensorFlow
• Python Packages: Matplotlib, Pandas, Numpy, Scikit-Learn, CUDA
• Version Control: Git, Github

Key Skills Demonstrated:

• Protein Modeling and Simulation
• Deep Learning (CNN, NLP)
• Structural Bioinformatics
• Sequence Analysis
• Python and C++ Programming
• High-Performance Computing (CUDA)

Project: Quantum Chemical Study of Carbon Ring Interactions with Carbon Nanotubes and Graphene

Summary:
Conducted a computational study employing quantum chemical theories to investigate the energetics of interactions between carbon rings and carbon nanotubes/graphene, shedding light on the fundamental mechanisms governing these systems.

Key Contributions:

• Quantum Chemical Simulations: Utilized Schrödinger's Material Science Suite to perform energy optimization and analyze the interactions between carbon rings and carbon nanotubes/graphene.
• Research Findings: Contributed to the understanding of the energetic and structural properties of these systems, providing valuable insights for future applications.
• Peer-Reviewed Publication: Co-authored a publication in ACS Omega, a renowned scientific journal, detailing the research methodology and findings.

Publication Details:

• Title: Adsorption of Monocyclic Carbon Rings on Graphene.
• DOI: 10.1021/acsomega.8b00378
• Journal: ACS Omega
• Year: 2018

Technical Skills Demonstrated:

• Quantum Chemical Simulations
• Energy Optimization
• Materials Science
• Computational Chemistry
• Schrödinger's Material Science Suite