hi, i'm sidhanth

Blame is a native macOS menu-bar app for developers running AI coding agents on 8–16 GB machines. It answers one question—why is my Mac slow right now, and what is safe to end?—by naming the cause: agent-session leftovers, forgotten dev servers, hidden automation browsers, VMs holding half the RAM, browsers open for days. Every fix passes a safety gate over the whole process tree, so databases, builds, and git are never touched. Diagnosis and one-click fixes are free; Autopilot, which acts on its own with a 10-minute undo, is $29 once.

Persnally is a personal context engine that learns from your AI history so every tool you use understands you—without re-explaining yourself. A local daemon synthesizes a structured model of who you are and serves it to Claude, Cursor, ChatGPT, and more over MCP; every claim cites its evidence, and any AI can call persnally_ask to get how you'd decide, with a confidence score. It lives in ~/.persnally on your machine—local-first, works offline, and truly deletable. Not our cloud. Yours.

Reminders is for groceries; Pingback is for the messages you owe a human. Forward a message from any app, speak it to Siri, or tap the lock-screen widget and Pingback files the follow-up in two seconds. Every morning at 7am it reads your open loops aloud while you make coffee. On-device first, no transcripts kept. iOS, free at launch.

Ravox is an audiobook production studio built for authors. Upload a manuscript and it analyzes chapters, characters, and dialogue, casts character-specific voices, and directs the performance with an Audio Direction Engine that controls emotion, pacing, and emphasis. Pronunciation Intelligence and Series Memory keep names, voices, and tone consistent across an entire series, while a line-by-line editing suite (trim, re-record, retune) exports ACX-ready audio mastered to Audible, Apple Books, and Spotify specs.

0byte stamps every AI generation with a fingerprint and a public record—who made it, with which model, and when. A perceptual fingerprint survives screenshots, crops, and re-encoding, while an Ed25519 signature and an append-only public ledger make each generation verifiable by anyone—a single SDK call, no account needed. Stamping runs in under 50ms in the background and leaves the content byte-for-byte identical, working across any AI model, including open source. Built for model providers, newsrooms, developer platforms, and creators who need to prove provenance.

Modelia is an AI platform that allows taking the basic images used by retailers to sell clothes and easily, quickly, and simply transforming them into high-quality images featuring AI-generated people of any age, gender, race, and size. Just by dragging the image of a garment, describing the desired model in text, pressing the generate image button, and downloading it for use on any platform.

A deep-tech company in Gurugram building the trust layer and offline-first tools for the AI-native internet. We fuse applied cryptography, blockchain rails, and on-device ML to ship products like 0byte (cryptographic provenance for AI-generated media) and Ravox (offline audiobook creation). We pair rigorous engineering with fast, product-first iteration—ship, learn, refine—to turn ambitious ideas into production-ready systems that feel simple, help people move faster, and let them trust what they see.

Increased user engagement by 30% through the development of AI-powered features, including GPT Plus model integration for PDF-based question generation. Reduced critical bug reports by 40% by implementing a comprehensive error tracking and resolution system, significantly improving overall user experience. Improved team productivity by 20% through the development of an integration between question segment reporting and Slack, optimizing workflow efficiency.