[Currently working on a native mobile app]
I thrive in my flow state when building AI-driven projects, exploring documentation, and debugging complex problems. I stay ahead of emerging technologies through blogs, articles, and in-depth tutorials. Check out my AI-focused certifications HERE.
MODULES: Stats for Data Science // ML, AI & Decision Sciences // Computer Vision & Deep Learning // Big Data
MODULES: Data Structures // C, C++, Python & Java // Security // IoT // Business Intelligence // Linux Admin
> AGENTIC AI & LLMs: LANGCHAIN // CREWAI // AUTOGEN // RAG // GEMINI // LLAMA
> AI & VISION: PYTORCH // TENSORFLOW // OPENCV // CNNs
> FULL STACK & OPS: MERN // MLFLOW // DOCKER // FASTAPI
Dec 2024 - Present
• • Building a cross-platform mobile app for mental health
professionals using React Native and MERN Stack.
• Engineered a secure Text-to-SQL agent using LangGraph
and Llama 3 via Groq for Azure SQL queries.
Jun 2024 - Dec 2024
• Developed a unified chatbot with RAG architecture using
Mixtral-8x7B LLM and FAISS vector store.
• Integrated speech-driven functionality using Google's
speech recognition and TTS packages.
Architected a multi-agent recommendation engine using CrewAI, Gemini, and MLflow for LLM observability.
> STACK: CREWAI // GEMINI // MLFLOW // MCP
// View DemoDeployed a multimodal agent system using AutoGen and Gemini with a Dockerized CI/CD pipeline.
> STACK: AUTOGEN // GEMINI // DOCKER // GITHUB CI/CD // AGENTOPS
// View DemoReplicates ChatGPT functionality using LangChain for conversational AI and Streamlit for the UI.
> STACK: LANGCHAIN // STREAMLIT // OPENAI_API
// View SourceUtilizes Hugging Face's LLAMA2 to assist users in effortlessly generating email content via Streamlit.
> STACK: LLAMA-2 // STREAMLIT // HUGGING_FACE
// View SourceComparative study using ensemble CNN models and ML classification for accurate detection of coffee diseases.
> STACK: CNN // ENSEMBLE_MODELS // PYTHON // ML
// View SourceDeveloped a ResNet classifier (98% accuracy) for tumor detection and ResUNet architecture for precise localization using LGG-MRI datasets.
> STACK: RESNET // RESUNET // TENSORFLOW // KERAS
// View SourceEngineered computer vision models for facial keypoints (83.9% accuracy) and emotion detection (88.2% accuracy) with robust data augmentation.
> STACK: OPENCV // CNN // TENSORFLOW // PYTHON
// View SourceLeverages GFP-GAN architecture to restore color accuracy, vibrancy, and sharpness to damaged vintage photographs.
> STACK: GFP-GAN // PYTORCH // GENERATIVE_AI
// View SourceAcademic social networking app built with React Native (Frontend), Node.js, and MongoDB.
> STACK: REACT_NATIVE // NODE.JS // MONGODB // EXPRESS JS
// View Source[Email: aishwaryapirankar@gmail.com]
[Github: github.com/aishwaryapirankar]
[LinkedIn: linkedin.com/in/aishwarya-pirankar]