Back

Raju Biswal

🧠 Building Laryaa a... • 2m

What if AI had to work on legacy systems, with no APIs, no data egress, and weak hardware? In many real-world enterprise environments, especially across emerging markets, AI systems face constraints that most modern agents are not designed for. There are no APIs. Sensitive data cannot leave the device. Applications are legacy Windows software. Hardware is often low-spec. Systems are shared and tightly regulated. Under these conditions, most AI agents fail — either technically or from a security and compliance standpoint. The core insight I arrived at was that planning, perception, and execution must be separated, both technically and trust-wise. Planning does not require raw data. Perception does not require cloud intelligence. Execution must remain local and deterministic. This led to a split-state architecture, where on-device vision interprets screen state, a local sanitization layer abstracts sensitive information, and only high-level intent is used for planning. All execution — and all sensitive data — remains on the device. The cloud never sees raw screens, identities, or records. To validate this, I built an early prototype system, now called Laryaa aOS. It is an execution platform designed to let AI agents operate inside legacy and regulated environments without APIs or data egress. The architecture was novel enough that a provisional patent was filed around the core execution and sanitization model. This is not an RPA tool, a vertical SaaS product, or a copilot. It is an execution layer for autonomous agents, with the hospital workflow serving only as one proof point. --- Laryaa has recently evolved into Laryaa aOS — not a single product, but a platform. Laryaa aOS is still early. We’re building, learning fast, and validating through conversations. At this stage, I’m: Exploring conversations with a co-founder (tech / product / GTM) Open to early investment discussions (pre-seed) For more details on the architecture, visit https://laryaa.com — Raju Founder, Laryaa aOS hello@laryaa.com

1 Reply
8
1
Replies (1)

More like this

Recommendations from Medial

Image Description
Image Description

Rahul Agarwal

Founder | Agentic AI... • 2m

If you use AI agents, you’ve probably faced these problems. I've explained 5 common issues and how to fix them. 1. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗟𝗼𝘀𝘀 𝗶𝗻 𝗖𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻𝘀 𝘊𝘩𝘢𝘭𝘭𝘦𝘯𝘨𝘦: AI agents often forget earlier parts of a conversation or

See More
1 Reply
4
8
1

Vansh Khandelwal

Full Stack Web Devel... • 6m

In the rapidly evolving digital landscape, APIs (Application Programming Interfaces) serve as essential connectors among software applications, enabling seamless communication between systems. However, their accessibility also makes them prime target

See More
Reply
3
Image Description
Image Description

Vikas Acharya

Building Reviv | Ent... • 1y

WHAT ARE AI AGENTS ? AI agents are software programs that use artificial intelligence to interact with their environment, collect data, and perform tasks. They can be programmed to achieve specific goals, and they can learn from their experiences.

See More
5 Replies
4
17

Raju Biswal

🧠 Building Laryaa a... • 2m

Why AI Automation Breaks in Regulated Environments? Most AI automation tools are designed for cloud-first, API-rich systems. But large parts of the real world don’t operate that way. This insight came directly from trying to deploy AI co-pilots i

See More
Reply
6

Abdul Shaikh

Every Dream is Worth... • 11m

🚫 Al Agents Are Coming-90% Will Fail Without This Key Factor Al agents promise seamless automation and intelligent decision-making, but their effectiveness hinges on one crucial factor: high-quality data. Without clean, structured, and accessible

See More
Reply
2
17

Devak K

Hey I am on Medial • 11m

Types of Data Breaches in Cyber Security | Digitdefence Understanding the types of data breaches in cyber security is crucial to protecting your business. Knowing these breach methods helps you implement stronger defenses and safeguard sensitive inf

See More
Reply
1

Rahul Agarwal

Founder | Agentic AI... • 3m

Everyone is building agents, but only a small number truly understand what it takes to run them in production. Here’s the part people usually avoid saying: 𝗠𝗮𝗸𝗶𝗻𝗴 𝗮𝗴𝗲𝗻𝘁𝘀 𝗶𝗻 𝗱𝗲𝘃 𝗶𝘀 𝗽𝗿𝗲𝘁𝘁𝘆 𝗲𝗮𝘀𝘆. In development, everything

See More
Reply
8

Anup Thatal

IT enthusiastic | Fu... • 7m

👋 Hello Startups! IT Engineer Seeking Exciting Opportunities in India 🚀 Hey everyone! I’m Anup Thatal, a software & data engineer with over 3 years of hands-on experience building scalable web apps, APIs, and digital infrastructure — especially in

See More
Reply
2
Image Description
Image Description

Rahul Agarwal

Founder | Agentic AI... • 2m

Data scientist, Data analyst, AI engineer, or AI agent builder? Which one is best? I've explained below. 1. 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 This field teaches you how to 𝗮𝗻𝗮𝗹𝘆𝘇𝗲 𝗱𝗮𝘁𝗮, 𝗯𝘂𝗶𝗹𝗱 𝗠𝗟 𝗺𝗼𝗱𝗲𝗹𝘀, 𝗮𝗻𝗱 𝗱𝗲𝗽𝗹𝗼𝘆 𝘁𝗵𝗲𝗺 𝗶

See More
1 Reply
22
20
2

Gigaversity

Gigaversity.in • 9m

What is an API? Have you ever ordered food at a restaurant? The internet works in a similar way! An API (Application Programming Interface) is like a waiter - it takes your request, delivers it to the kitchen (backend), and brings back the response.

See More
Reply
7

Download the medial app to read full posts, comements and news.