Age of Cloud AI
Modern artificial intelligence has developed an uncomfortable dependency: almost every interaction we have with a language model requires a round-trip ticket to a remote server. We ask intimate personal questions, draft sensitive documents, brainstorm business strategies, and explore our raw thoughts through an interface that silently transmits our keystrokes to centralized data centers. While technology companies offer terms of service assuring users that data will remain private, the architectural reality remains unchanged: if your data leaves your physical device, your ownership of it becomes conditional. A promise of privacy is never a substitute for architectural privacy.
Shubhrali was conceived to challenge that paradigm. The objective was straightforward yet technically demanding: build an artificial intelligence companion that lives entirely inside the user’s pocket, executing offline without sending a single byte of personal telemetry to any server on Earth. With recent breakthroughs in small language model efficiency and quantized tensor computation, we no longer need massive warehouse clusters to achieve competent, nuanced language understanding. By pairing high-density models like Qwen 2.5 with efficient 4-bit quantization and mobile C++ inference runtimes, Shubhrali proves that powerful reasoning and high-speed token generation can run comfortably on standard, everyday mobile silicon.
The core motivation behind Shubhrali is not merely technical curiosity; it is digital sovereignty. True privacy is not an encrypted toggle in a settings menu managed by an external company—it is the physical impossibility of data extraction. When you run an inference pipeline locally in your phone's memory, your thoughts remain yours. You can converse freely in airplane mode, in regions with zero network infrastructure, or in environments where confidentiality is paramount. Shubhrali eliminates subscription gatekeeping, server downtime, and corporate surveillance by returning computational control to the individual.
Building this platform has meant confronting real edge-computing constraints—balancing memory footprints to prevent low-memory kills on devices with four gigabytes of RAM, optimizing memory-mapped model weights, and ensuring stable token generation speeds across diverse mobile chipsets. Yet every performance optimization reinforces our core belief: intelligence does not belong exclusively behind a paywalled, cloud-hosted API. Shubhrali exists to make private, deterministic, and locally owned AI accessible to anyone holding a phone. This project is a return to computing as it was always meant to be—personal, self-contained, and completely free.
The Soul Behind Shubhrali
If you look at the artificial intelligence landscape today, almost everything looks, sounds, and feels identical. It is an endless sea of sterile corporate utilities, monochrome chat bubbles, and sanitized algorithms designed in boardrooms to be as inoffensive and forgettable as possible. They function like cold office appliances: efficient, distant, and completely devoid of spirit. But a tool without an identity cannot form a genuine relationship with the person using it. You cannot truly connect with people without a personality, a voice, and an uncompromising set of values of your own.
That belief is why nothing in Shubhrali is accidental. Every single design choice, from the name itself to the emblem on the interface, was chosen because it carries a specific weight and intention. The name Shubhrali draws from auspiciousness, clarity, and light—a deliberate departure from clinical tech acronyms and mechanical code numbers. It represents warmth, an intelligent companion that illuminates your thoughts rather than recording them. The constellation mark—anchored by the sign of Leo—embodies courage, fierce independence, and sovereignty. A lion does not ask permission from a distant authority to exist in its domain, and your thoughts should not have to ask permission from a corporate server to be processed.
People do not trust or remember faceless corporate entities; they connect with creations that have a heartbeat and stand for something unshakeable. When you open Shubhrali, you are not just interacting with quantized mathematical weights running on your phone's processor. You are stepping into an environment shaped by a clear philosophy: that personal computing must remain personal, that private dignity is non-negotiable, and that genuine intelligence must feel grounded and human.
This is the exact reason Shubhrali was engineered to run strictly offline on your own device. Modern cloud-based assistants make your privacy conditional, quietly packaging your questions and ideas into telemetry for external servers. Shubhrali rejects that model entirely. By confining every tensor calculation, every prompt, and every generated word inside your phone's physical memory, it turns technology back into what it was always meant to be: a self-contained instrument that answers only to you.
What remains is something rare in modern technology: an application with a soul. Shubhrali was built with meaning because the people who use it deserve more than another disposable utility. They deserve a companion built with courage, intention, and a clear set of values—one that stays by their side, keeps their trust completely intact, and speaks with a character entirely its own.
Terms of Use and Privacy, Medical and Legal Policy
PREAMBLE
This Master Legal and Privacy Agreement (hereinafter referred to as the "Agreement") constitutes a legally binding and enforceable contract entered into by and between the individual or entity accessing the Platform (hereinafter referred to as the "User", "you", or "your") and Shubhrali AI, an emerging technology initiative incubated, governed, and legally represented under Proton India Trust (hereinafter collectively referred to as the "Company", "we", "us", or "our").
Proton India Trust serves as the primary legal entity, administrative representative, and fiduciary guardian for Shubhrali AI in all official capacities, statutory filings, regulatory compliance operations, and judicial proceedings. By downloading, accessing, installing, compiling, registering with, or utilizing the Minimum Viable Product (MVP) application, client-side binaries, associated website, local machine learning models, or digital assets contained therein (collectively, the "Platform"), the User explicitly acknowledges, warrants, and confirms that they have read, understood, and consented to be bound without reservation by all terms, conditions, protocols, and disclaimers set forth in this Agreement.
ARTICLE I: PRIVACY POLICY AND DATA GOVERNANCE
Section 1.01. Local Data Processing Architecture
The Platform is engineered from the ground up as an offline-first, client-side intelligence system. All conversational inference, natural language generation, context compilation, prompt processing, and mathematical tensor evaluations are executed locally within the physical volatile memory (RAM) and processor cache of the User’s local computing device. The Company does not host, intercept, collect, monitor, harvest, log, or store the User's chat history, uploaded prompt data, behavioral profiles, emotional states, or conversational context on any centralized server, cloud repository, or third-party infrastructure.
Section 1.02. Online Mode and Zero-PII Web Grounding
In scenarios where the User's computing device maintains an active internet connection, the Platform’s auxiliary search modules may execute client-initiated web queries to retrieve current factual reference data. All outbound queries generated in this operational mode are strictly stripped of all Personally Identifiable Information (PII), device identifiers, session fingerprints, and unique user tokens before transmission. The retrieved raw search results are piped directly back to the local client, where all contextual evaluations, summarizations, and response generations take place entirely on the local device hardware.
Section 1.03. Account Authentication and Licensing Telemetry
Access to certain distribution channels or features of the Platform may necessitate one-time identity verification via a valid Google Account or OAuth token. This authentication layer functions strictly for administrative abuse prevention, bot protection, and license seat management. Identity verification metadata is isolated completely from model inference runtimes. The Company under no circumstances links, associates, correlates, or cross-references the User’s authenticated Google profile with the contents of their local model queries, offline logs, or local device storage.
Section 1.04. System Updates and Model Downloads
The Company may deploy updates to improve application components, including user interface wrappers, runtime engines, and quantized small language model binaries. Such updates operate unidirectionally (server-to-client) and contain zero telemetry payloads capable of inspecting, scraping, or transmitting files from the User’s file system.
Section 1.05. Age Restrictions and Minor Protection
The Platform is intended exclusively for access and utilization by individuals aged thirteen (13) years or older. The Company does not knowingly register, collect data from, or target individuals under the age of thirteen. If the Company becomes aware that an individual under the age of thirteen has established an authenticated profile, the Company shall take immediate steps to terminate the linked authentication credentials and purge associated records from its authentication servers.
Section 1.06. Data Deletion and Right to be Forgotten
Users retain the absolute right to request the permanent revocation, unlinking, and deletion of their authenticated Google Account records and licensing metadata from the Company’s authentication infrastructure. Such requests shall be executed upon receipt of a formal request submitted to the official electronic mail address specified in Article VIII of this Agreement.
ARTICLE II: COOKIES AND METADATA POLICY
Section 2.01. Indexing, Structured Data, and Web Crawlers
The website associated with the Platform employs industry-standard Open Graph tags, JSON-LD schemas, and structured metadata solely to facilitate accurate content categorization and indexing by global search engines and automated conversational crawlers (including, without limitation, Googlebot, Bingbot, Grok, and Gemini).
Section 2.02. Strictly Necessary Operational Cookies
The Platform’s web interfaces employ strictly necessary, first-party operational cookies and session tokens essential to maintain site stability, CSRF security, asset caching, and basic interface responsiveness. Navigating the Platform’s web domain constitutes explicit consent to the placement of these functional tokens. The Company deploys no third-party behavioral advertising trackers, ad-network tracking pixels, or cross-site data harvesting scripts.
ARTICLE III: MEDICAL, PSYCHOLOGICAL, AND CRISIS DISCLAIMER
Section 3.01. Absolute Limitation of Medical and Mental Health Advice
The Platform functions solely as an assistive, local software utility and empathetic conversational text companion. The Platform, its underlying language models, its developers, and Proton India Trust are not licensed medical professionals, clinical therapists, psychiatric practitioners, or emergency healthcare services. The Platform lacks the professional, scientific, and diagnostic competence to evaluate, diagnose, treat, prevent, or cure any clinical psychiatric disorders, psychological pathologies, physiological illnesses, or mental health emergencies.
Section 3.02. Emergency and Crisis Intervention Protocols
Under circumstances where the Platform's local language processing system detects patterns indicative of severe mental health crises, acute personal distress, clinical depression, or explicit intent of self-harm or suicide, the software is algorithmically instructed to trigger an immediate, high-priority intervention prompt. This safety mechanism displays actionable references to regional emergency crisis resources, specifically prioritizing emergency helplines within the Republic of India, including 112 (National Emergency Support) and 108 (Emergency Medical Services).
Section 3.03. Manual Action Requirement and Limitation of Liability
The User explicitly acknowledges that the Platform possesses no hardware capacity, programmatic authorization, or telecom privilege to automatically dial emergency services, ping geographical location coordinates, or dispatch first responders. Initiating emergency intervention remains the sole responsibility of the User or their legal guardians. The Company, its trustees, developers, and governing entities disclaim all civil and criminal liability arising from the software's misinterpretation of emotional states, algorithmic hallucinations, delays in user-side emergency actions, or failure to secure external intervention.
ARTICLE IV: INTELLECTUAL PROPERTY, COPYRIGHT, AND CORPORATE OWNERSHIP
Section 4.01. Proprietary Assets and Corporate Ownership
All proprietary software components, graphical user interfaces, design architecture, system styling, website source code, typography styling, custom algorithms, prompt engineering systems, documentation, and original visual marks—specifically including the proprietary modified Leo/Simha constellation iconography—are the sole, 100% exclusive intellectual property of Shubhrali AI. No individual developer, contractor, or founding contributor asserts personal ownership rights over the collective core assets of the Platform.
Section 4.02. Strict Copyright Reservation
All proprietary aspects of the Platform are protected under the provisions of the Indian Copyright Act, 1957, international copyright conventions, and multilateral intellectual property treaties. Any unauthorized replication, decompilation of proprietary binaries, systematic scraping, mirrored republication, or unauthorized commercial exploitation of Shubhrali AI’s brand assets, written copy, or distinctive interface elements without prior express written consent is strictly prohibited.
Section 4.03. Open-Source Dependencies and Model Attributions
The Platform utilizes and interoperates with third-party open-weight foundational language models, including but not limited to the Qwen model series (governed by Alibaba Cloud’s open-weight licenses) and Falcon model architectures. Furthermore, the MVP client application shell incorporates components derived from Pocketpal AI, distributed under the permissive terms of the Apache License, Version 2.0. The User agrees to comply with all upstream open-source license provisions, copyright notices, and usage restrictions mandated by these underlying dependencies.
Section 4.04. Statutory Enforcement and Legal Recourse
The Company actively audits and monitors the public domain for unauthorized distributions and infringement of its proprietary assets. Any individual or corporate entity found pirating, reverse-compiling proprietary release builds, distributing trojanized clones, or falsely passing off Shubhrali AI’s marks shall be prosecuted to the fullest extent of civil and criminal intellectual property law.
ARTICLE V: ACCEPTABLE USE AND USER CONDUCT
Section 5.01. Prohibited Uses and Unlawful Activities
The User agrees that they shall not utilize the Platform, directly or indirectly, to draft, facilitate, compile, or distribute:
- Malicious cybersecurity payloads, exploits, denial-of-service scripts, keyloggers, or unauthorized infiltration utilities;
- Unlawful, explicitly abusive, or non-consensual sexually explicit (NSFW) materials;
- Instructions, chemical formulations, or kinetic planning intended to inflict physical violence, bodily harm, terror attacks, or unlawful harm against any person, group, or infrastructure;
- Defamatory campaigns, targeted harassment, stalking, or fraudulent financial impersonation materials.
Section 5.02. Ethical Rejection and Account Termination
The Platform’s local instruction models are aligned to refuse unethical, destructive, or illicit directives. Repeated attempts to bypass the model’s ethical boundaries via adversarial jailbreaks, prompt injection, or malicious payload compilation shall constitute an immediate material breach of this Agreement. The Company reserves the unilateral authority to terminate authentication privileges, void testing licenses, and block access tokens associated with offending Google accounts.
ARTICLE VI: END-USER LICENSE AGREEMENT (EULA) AND COMMERCIAL TERMS
Section 6.01. Limited MVP Evaluation License
With respect exclusively to the current Minimum Viable Product (MVP) release, the Company grants the User a personal, revocable, non-exclusive, non-transferable, and royalty-free license to download, install, and execute the application binaries on personal hardware. The User is permitted to evaluate, test, inspect, and benchmark the runtime solely for non-commercial educational, testing, and feedback purposes. This evaluation license conveys no transfer of copyright, patent interest, or brand ownership to the User.
Section 6.02. Future Commercial Iterations and Monetization
The MVP release is offered free of charge for evaluation purposes. The Company expressly reserves the unilateral right to introduce commercial licensing tiers, enterprise deployment fees, subscription models, or proprietary feature paywalls in subsequent stable releases and enterprise distributions without prior individual notice.
ARTICLE VII: GOVERNING LAW AND DISPUTE RESOLUTION
Section 7.01. Applicable Governing Law
This Agreement, its construction, its validity, and all legal relationships, obligations, and disputes arising out of or connected with the Platform shall be governed exclusively by, interpreted, and enforced in accordance with the substantive and procedural laws of the Republic of India, without reference to its conflict-of-law principles.
Section 7.02. Exclusive Judicial Venue and Jurisdiction
The User irrevocably submits and agrees that any action, suit, arbitration, legal proceeding, or litigation instituted by either party to enforce, interpret, or resolve disputes under this Agreement shall be brought exclusively before the competent courts of civil and criminal jurisdiction situated in Lucknow, Uttar Pradesh, India. The User specifically and unconditionally consents to the personal and territorial jurisdiction of the High Court of Judicature at Allahabad, Lucknow Bench, and its subordinate courts in Lucknow, waiving any defense based upon an inconvenient forum.
ARTICLE VIII: OFFICIAL CONTACT AND STATUTORY COMMUNICATIONS
Section 8.01. Official Inquiries and Notice Pipeline
For all formal legal notices, data unlinking and deletion requests, copyright infringement notifications, regulatory communications, and technical security disclosures, you may contact the Company's designated administrative channel via electronic mail at:
Email: shubhrali.ind@gmail.com
Governing Authority: Proton India Trust (Legal Custodian for Shubhrali AI)
Location: Lucknow, Uttar Pradesh, India
Is the Founder Eligible?
In the technology industry, a prevailing assumption suggests that consequential advances in applied artificial intelligence belong exclusively to multi-billion-dollar corporate laboratories or enterprise research divisions. When an independent developer engineers an on-device, privacy-first computing platform to challenge conventional cloud dependency, questions of capability, credentials, and domain eligibility naturally follow.
Eligibility in software engineering is neither bestowed by corporate tenure nor granted by corporate permission. It is established through technical competence, execution across resource-constrained systems, and alignment with principled engineering standards.
1. Structural Autonomy from Cloud-Centric Business Models
Large technology conglomerates operate under fundamental economic incentives that favor centralized data harvesting, recurring subscription services, and compute telemetry. Engineers operating within these environments are bound to architectures that sustain recurring cloud infrastructure margins and corporate data pipelines. Consequently, the creation of a truly client-side, air-gapped intelligence platform contradicts the core commercial mandates of traditional cloud enterprises.
Building Shubhrali independently eliminates these conflicts of interest. The platform is not subject to corporate monetization pressures, telemetry targets, or recurring token rent. Operating outside corporate constraints allows architectural decisions to remain focused entirely on user privacy, computational efficiency, and client-side data sovereignty.
2. Comprehensive Systems Engineering and Technical Competence
Deploying localized Large Language Models on consumer-grade mobile hardware requires end-to-end systems programming, spanning low-level operating system bindings to mathematical optimization:
- Applied Tensor Optimization: Quantizing high-dimensional transformer checkpoints requires deep technical fluency in tensor mechanics. Shubhrali leverages medium k-quantization (Q4_K_M) and Grouped-Query Attention (GQA) to maintain low perplexity metrics while compressing parameter footprints to practical mobile dimensions.
- Low-Level Native Execution: Running real-time inference on edge silicon demands optimization below the application layer. The platform relies on native C++ execution environments, direct memory mapping (mmap), dynamic cache management, and ARM NEON SIMD vectorization to extract peak computational throughput from mobile chipsets.
- Hardware Memory and Thermal Calibration: Developing for standard mobile hardware requires managing strict OS memory ceilings. Calibrating key-value cache states and context buffers prevents low-memory termination (OOM) on systems with limited RAM, ensuring stable and thermally balanced execution.
Mastering this stack demonstrates that eligibility is defined by hands-on technical execution and the ability to solve systems-level edge computing constraints directly.
3. Purpose-Driven Engineering and Identity
Modern enterprise software frequently prioritizes sterile, interchangeable utilities that treat users as passive nodes within a distributed network. Shubhrali was deliberately designed around an explicit identity and clear values:
- Symbolism and Purpose: The name Shubhrali draws from themes of clarity, auspiciousness, and illumination. The accompanying constellation emblem, inspired by Leo, reflects autonomy, courage, and self-determination.
- Human-Centered Privacy: Technology must respect human dignity. When users interact with a computational assistant, their private reflections, strategies, and personal data should remain strictly confidential. Shubhrali translates this philosophy into a deterministic hardware reality: if data never leaves physical device memory, it can never be intercepted or compromised.
4. Incubation, Governance, and Structural Accountability
While Shubhrali is driven by dedicated technical leadership, it does not function without formal oversight. The initiative is officially incubated under Proton India Trust, providing administrative structure, strategic advisory, and foundational governance. This framework ensures that independent technical innovation is paired with long-term accountability, legal integrity, and a public-service mission.
Historical precedent in computing consistently demonstrates that paradigm shifts originate from focused individuals willing to challenge prevailing industry assumptions from first principles. Shubhrali was developed with the technical competence to execute edge AI, formal governance to ensure accountability, and the philosophical conviction that intelligence must remain personal, sovereign, and private.
