Founder & CEO at Magnence. AI Engineer & Software Architect

productHunt

github

Anurag Kumar Singh • Peerlist

<h3>A Comprehensive Suite of Browser-Based Digital Tools</h3>ToolLoom offers a powerful and privacy-focused platform with over 250 client-side tools designed to handle a wide range of file processing needs. From PDF manipulation and image editing to video transcoding and developer utilities, all operations are performed directly within your local browser using WebAssembly. This ensures that your files never leave your device, providing complete privacy and eliminating the need for cloud uploads.<ul><li>Extensive Toolset: Access a vast collection of tools categorized for easy navigation, including PDF management, image processing, file conversion, video and audio editing, developer utilities, office document handling, archive management, text manipulation, and social media presets.</li><li>100% Client-Side Privacy: All file processing occurs locally in your browser's sandbox. Your sensitive data remains on your device, ensuring compliance with GDPR, HIPAA, and corporate NDA requirements.</li><li>High Performance: Leveraging WebAssembly, ToolLoom delivers near-native processing speeds for demanding tasks like video compression and batch file operations, utilizing your device's local CPU and GPU power.</li><li>Offline Functionality: Once loaded, ToolLoom tools can operate offline, making them ideal for use on airplanes or in environments with limited internet connectivity.</li><li>No Limits or Registrations: Enjoy frictionless access with no account requirements, subscriptions, file size limits, or speed throttles. Capacity is determined solely by your device's resources.</li></ul>ToolLoom is built with modern web technologies to provide a seamless and efficient user experience without compromising on privacy or performance.

ToolLoom — Every Tool. One Place.

Built ApplyOps, an AI-powered job search and application management platform that brings job discovery, AI-based job matching, resume tailoring, and application tracking into a unified workflow. Designed an orchestration layer that coordinates multiple job-board extractors, processes job listings through AI pipelines, and synchronizes results with a real-time application tracking dashboard. Implemented AI-powered job-fit scoring to evaluate job descriptions against candidate profiles, along with sponsorship matching to help identify relevant opportunities based on work-authorization requirements. Developed one-click CV tailoring that analyzes job descriptions and generates targeted resume content, with automated PDF generation through LaTeX/Typst templates. Built automated Gmail integration to analyze recruiter responses and update application stages such as Interviewing and Rejected, reducing manual tracking throughout the job-search lifecycle. Designed the application as a modular, self-hostable system using React, TypeScript, Express, SQLite, Drizzle ORM, Docker, and multiple LLM providers including OpenAI, Gemini, OpenRouter, and local LLM endpoints. Tech: AI/LLMs · Python · TypeScript · React · Express · SQLite · Drizzle ORM · Docker · Web Scraping · REST APIs · Gmail OAuth · Automation · System Architecture

ApplyOps — AI-Powered Job Search & Applicatio

Built an end-to-end LLM pipeline from scratch in PyTorch, implementing core GPT/Transformer components including tokenization, embeddings, LayerNorm, GELU feed-forward networks, causal multi-head self-attention, Transformer blocks, and autoregressive text generation. Developed and trained GPT-style models, implemented temperature sampling and top-k decoding, and explored instruction fine-tuning using GPT-2 style architectures. Added checkpoint saving and loading for locally trained models. Built a Streamlit-based local inference interface that automatically detects model configurations from `.pth` checkpoints and provides configurable generation controls such as temperature, top-k, and maximum new tokens. The project demonstrates hands-on understanding of Transformer architecture, attention mechanisms, LLM training, instruction tuning, model inference, and practical deployment of locally trained language models. Tech: Python · PyTorch · Transformers · GPT · LLMs · Deep Learning · Streamlit · Model Fine-Tuning

LLM Creation From Scratch & Fine-Tuning

Built SelfGPT Studio, a local-first, self-hosted AI IDE that uses multiple LLM providers to generate, modify, debug, and run complete multi-file full-stack applications directly within a user's workspace. Designed an agent-based development workflow with specialized Builder, Debugger, Refactor, UX, Performance, Security, Test, and Documentation agents, enabling AI-assisted development across the software lifecycle. Architected the platform with Next.js 14, FastAPI, PostgreSQL, and local project storage, with support for multiple application stacks including React, Next.js, Node.js, MERN, Python, FastAPI, Flask, Django, and static web applications. Implemented project snapshots, file-level editing, diff comparison, dependency management, terminal execution, project export/import, stack-aware run profiles, and local application previews. Integrated multiple LLM providers including Groq, OpenAI, Mistral, DeepSeek, and Gemini, creating a flexible multi-model architecture for AI-powered software development. Tech: Next.js · FastAPI · Python · PostgreSQL · AI Agents · LLMs · Generative AI · Multi-Agent Systems · REST APIs · Docker · Full-Stack Development

SelfGPT Studio — Local-First AI IDE

Built UrbanMind, a real-time urban traffic intelligence platform combining AI-powered computer vision, adaptive traffic management, emergency response automation, and digital-twin visualization into a unified mobility system. Developed YOLOv8-based vehicle detection for traffic footage analysis, including vehicle classification, lane-density analysis, detection confidence, annotated evidence frames, and signal-timing recommendations based on traffic conditions. Designed a real-time architecture using FastAPI, React, Redis, WebSockets, and Docker to synchronize traffic data across dashboards, analytics, emergency operations, ROI analysis, and digital-twin interfaces. Implemented an emergency response workflow that tracks ambulance, fire, and police units using GPS telemetry, activates green corridors, identifies congestion along active routes, and generates alternate-corridor recommendations. Built operator-facing modules for live multi-sector monitoring, manual vision analysis, emergency dispatch, traffic analytics, and spatial visualization across a simulated 9-sector urban traffic network. Tech: Python · FastAPI · React · TypeScript · YOLOv8 · Computer Vision · OpenCV · Redis · WebSockets · Docker · Leaflet · Zustand · Three.js