Just a chill guy...

github

Arittra Bag • Peerlist

A genuine thank you to all the peers who supported and shared feedback on the earlier <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://peerlist.io/arittrabag/project/opsgraph">OpsGraph</a> launch. That feedback helped shape a much better 1.0.<a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://opsgraph-site-seven.vercel.app/">OpsGraph 1.0</a> is an open-source, self-hosted workspace for evidence-first PostgreSQL investigations.Ask questions in natural language. Your model proposes SQL, while OpsGraph enforces approved tables, read-only access, query limits and timeouts. Inspect the queries and captured records behind each answer.1.0 brings guided no-code setup with an interactive TUI, practice data, follow-ups in one conversation and readable incident reports.Use Anthropic, OpenRouter, OpenAI or Ollama. Hosting guidance covers local PostgreSQL, Supabase, Neon, AWS RDS, Google Cloud SQL, Azure and DigitalOcean.Start with the <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://opsgraph-site-seven.vercel.app/#start">guided setup</a> or <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://github.com/Arittra-Bag/opsgraph/blob/main/docs/quickstart.md#no-code-setup">step-by-step instructions</a>. Prefer Docker? Get the <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://github.com/Arittra-Bag/opsgraph/pkgs/container/opsgraph">GHCR images</a> and follow the <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://github.com/Arittra-Bag/opsgraph/blob/main/docs/installation.md#containers">container setup guide</a>.See <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://github.com/Arittra-Bag/opsgraph/releases/tag/v1.0.0">what changed in 1.0.0</a> or explore the <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://github.com/Arittra-Bag/opsgraph">source on GitHub</a>.The scope remains PostgreSQL, one operator and read-only investigations.

OpsGraph 1.0

<h3>About the Project</h3>Dataset-Maker creates reproducible torn-document datasets with the answer key built in.Upload a PDF, choose how many fragments you want and how rough the tears should be, and Dataset-Maker splits each page into non-overlapping pieces while recording exactly where every fragment belongs and which pieces share a torn edge.The goal is simple: document-reassembly research needs torn documents where the correct solution is known. Physical tears do not come with ground truth, and ad-hoc generators are hard to compare or reproduce.What it guarantees<ul><li>Strict partitioning: every page pixel belongs to exactly one fragment, with no overlaps or gaps. Every generated page is verified before export.</li><li>Deterministic tears: the same seed, tear parameters and DPI reproduce the same partition, placement and adjacency.</li><li>Exact ground truth: every fragment's page position and dimensions, plus the list of fragment pairs that share a torn edge.</li><li>Reproducible provenance: `manifest.json` records seeds, generation parameters, the input SHA-256 and relevant library versions.</li><li>Inspectable output: the web app lets you inspect the source page, partition map, adjacency graph, fragments and manifest before you download the export.</li></ul>The tearing model starts from a Voronoi partition and domain-warps the coordinate field with value noise, creating irregular boundaries while preserving the strict partition guarantee.<h3>dm-bench</h3>Dataset-Maker also ships dm-bench 0.1.0, a versioned benchmark for torn-document reassembly with easy, medium and hard tiers, public val and test-dev splits, and a held-out test split generated from a private secret, with a published commitment so its answers cannot be changed later.It includes an evaluation harness for placement accuracy, adjacency F1, perfect pages and candidate ranking, plus edge-greedy@0.1, a deterministic classical baseline that reads only the puzzle fragments.On the held-out test split, the baseline reaches 0.993 direct accuracy on easy, 0.865 on medium and 0.246 on hard. It solves 0% of hard pages perfectly, despite Hit@1 remaining above 0.9, leaving a clear gap between identifying neighbouring pieces and reconstructing the full page.Benchmark pages are synthetic and generated from a seed, so there is no personal or copyrighted data; a real-scan set is future work.dm-bench 0.1.0 is published on Hugging Face with a DOI, and Dataset-Maker is open source under the MIT License.

Dataset-Maker

Track your Ethereum (ETH) investments and balances with this simple web-based cryptocurrency portfolio tracker. Features: Connect Your Wallet: Easily connect your Ethereum wallet (e.g., MetaMask) to view your account details and balances. Add Multiple Accounts: Keep track of investments across multiple Ethereum accounts. Portfolio Summary: See an overview of your account balances in Ether (ETH). PWA Support: Install the tracker on your device for quick access. Send Cryptocurrency: Send Ether (ETH) to other addresses directly from the tracker.

Cryptocurrency Portfolio Tracker

DataVizX is a web application that allows users to explore and visualize their data in an Interactive manner. With a sleek and modern user interface, DataVizX makes data visualization a breeze, whether you're a data analyst, a business professional, or just someone who loves playing with data.

DataVizX

Dementia Dreamscape is a modern web application designed to assist in the early detection of Alzheimer's disease and related dementia conditions using brain MRI analysis. It features a responsive, AI-powered interface that provides real-time classification, confidence scoring, and AI-generated insights. Built using React, TypeScript, and TailwindCSS on the frontend, and FastAPI with TensorFlow on the backend, the platform analyzes uploaded MRI scans and returns predictions with confidence scores and recommendations. It integrates Google Gemini for enhanced interpretability and provides educational content to support dementia awareness. This project is intended for educational and research use only and is not a substitute for professional medical diagnosis.

Dementia Dreamscape | AI-Powered MRI Dementia

Disease Predictor

Developed a machine learning model using XGBoost to predict LN_IC50 for cancer drug responses with an impressive R² score of 99.41%. Leveraged SHAP for model interpretability, providing clear insights into feature contributions. Deployed the solution using Gradio, enabling real-time predictions through an interactive web interface. This project aims to assist researchers and pharmaceutical companies in evaluating drug efficacy and advancing personalized cancer treatment strategies.

GDSC Drug Sensitivity Predictor

AI-powered cardiovascular disease risk predictor using XGBoost, with SHAP for explainability and Gradio for an interactive web app. Achieved 73.66% accuracy, leveraging key health indicators like blood pressure, cholesterol, and lifestyle factors.

Heart Disease Risk Predictor with XGBoost & S

MindScope is an AI-powered tool that predicts depression and anxiety severity using XGBoost models trained on mental health data. It features a Streamlit-based interface for users to input their details and receive instant predictions, enabling self-assessment and awareness.

MindScope: Self-Analysis Mental Health Model

This project provides an interactive interface to classify MRI scans for Alzheimer's disease and brain tumors using transformer-based SegFormer models (nvidia/mit-b1). Each model is fine-tuned for image classification with four output classes specific to the condition. The models were trained on labeled MRI data using data augmentation techniques to improve generalization and reduce overfitting. Both models achieved an average validation accuracy of approximately 99.5%. The system is deployed using Gradio, enabling users to upload an MRI scan, select the condition type, and receive an instant prediction.

MRI Classifier: Alzheimer’s & Tumor Detection

NeuraScan is a full-stack AI-driven web application designed to classify brain tumors from MRI scans in real time. The model is trained using EfficientNet and implemented with PyTorch, enabling accurate multi-class classification across tumor types such as glioma, meningioma, pituitary, and no tumor. The application features a FastAPI backend that handles image preprocessing, model inference, and secure API interactions. On the frontend, it uses React 18, TypeScript, and TailwindCSS to deliver a smooth, responsive user interface. Animations and UI elements are enhanced with GSAP, Lucide Icons, and Shadcn UI components. To enrich user experience with interpretability, the system integrates Google Gemini to generate AI-powered medical insights and explainable reports based on classification results. The platform emphasizes: 1. Real-time MRI image classification 2. Confidence scoring for each tumor type 3. AI-generated medical recommendations 4. Educational utility in healthcare research and pre-diagnostic support All data is processed in-memory with no storage, ensuring user privacy. The system is CORS-protected, API key–secured, and production-ready. This project demonstrates expertise in deep learning, medical imaging, model deployment, full-stack development, and real-time AI applications in healthcare.

NeuraScan | AI-Powered Brain Tumor Detection 

High-performance quantitative trading simulator leveraging real-time L2 orderbook data from OKX via WebSocket, with advanced models for slippage, market impact, and fee estimation. Integrated AI market sentiment analysis using Google's Gemini. Built with Python, Dash, and Plotly, featuring latency metrics, backtesting support, and modular architecture.

Quant Trade Simulator

TerraBoost

OpsGraph is an open-source, self-hosted workspace for evidence-first PostgreSQL investigations.Ask a question in natural language and let your configured model propose bounded queries. OpsGraph - not the model - controls what can actually execute: approved schemas and tables, read-only access, query count, row limits and timeouts.Every finding remains connected to the SQL and captured records behind it, so you can inspect the evidence instead of blindly trusting a generated answer.<a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://github.com/Arittra-Bag/opsgraph/releases/tag/v0.1.0b1">Beta v0.1.0b1</a> adds saved and reopenable investigations, linked follow-ups, evidence exports, local Ollama / OpenAI-compatible models, Anthropic support, native macOS/Linux/Windows installation and a multi-architecture Docker image.The scope is intentionally narrow for now: PostgreSQL, one operator, read-only investigations.