01
Custom AI agents and multi-agent workflows
Agents that carry a task from the request to a finished result. When the work splits cleanly, several agents divide it, and a person still approves the step that matters.
Software studio
Mycelium Tech designs and ships web apps, mobile apps, SaaS products, and custom software. Agents, voice, and models that read your own files are the part we go further on, from the first sketch to production.
Services
Two practices, one team. Product engineering for the thing people touch. AI engineering for the part that listens, decides, or answers.
AI engineering
01
Agents that carry a task from the request to a finished result. When the work splits cleanly, several agents divide it, and a person still approves the step that matters.
02
Real-time, low-latency phone agents on LiveKit and Twilio. Built for calls that cannot drag: restaurants taking reservations and orders, law firms covering intake, and taxi services booking and updating rides.
03
Products that answer from your documents. Retrieval, citations, and the decision to decline an unsupported answer are part of the build.
04
Private AWS deployments, including local models, for teams whose data cannot leave their network.
05
The repetitive movement of data between tools, replaced with a flow your staff can see, pause, and trust.
06
A narrow version of an AI product, in production, so you can learn from real use before you fund the rest of it.
Product engineering
01
Product sites, customer portals, and internal web apps with the performance, accessibility, and structure a team can keep owning.
02
iOS and Android products with deliberate flows, a release path, and an interface that matches the web product beside them.
03
Multi-tenant products with accounts, roles, billing, and the admin tools a real customer base requires.
04
Systems shaped around one operation: the handoffs, permissions, and data your team already lives with.
05
Interface design done with the people who will build it, so the screens you approve are the screens that ship.
Use cases
The same studio builds the product and the system behind a phone call, a document, or a private deployment.
A phone agent that takes reservations and orders, and hands the messy exception to a person.
After-hours and overflow intake that records who is calling and what the matter is about, then routes a summary. It does not pretend to be the attorney.
Call handling that books a trip, updates it, and passes the details to dispatch.
Web and mobile products, from a first release to the version a customer can pay for.
Workflow automation across the systems people already have open all day.
Assistants and models running in your own AWS account when the data has to stay there.
How we work
We learn the user, the constraint, and what done means. You leave with a scope you can accept or refuse.
A working slice you can click, call, or hand to a colleague. Enough to make a decision, and built so it can continue.
Design, engineering, and the parts that keep it alive: auth, data, logging, and a deploy you own.
We stay after launch for fixes and the next increment, with the context of the people who built it.
Capabilities
A system is easier to picture than a slogan. These are the kinds of work we design and ship.
Voice
Voice
LLM
Product
Product and voice
Infrastructure
Capability tiles, not client stories. Named projects are on the work page.
Selected work
Lead projects from the studio. The full list, including work still in progress, is on the portfolio page.
Stack
A palette, not a promise that every project uses all of them.
Hosted models when a frontier model fits the product and the data policy allows it.
When we need to choose the model, or run it ourselves.
The real-time path for phone agents.
Private deployments, and the rest of the infrastructure around them.
Application data and hosting when they are the right fit.
The languages the studio builds in.
The web interface layer.
Studio
Mycelium Tech is led by Rizwan Mughal, Senior AI Software Architect. He has spent six years on enterprise agentic AI and voice systems, and he stays close to the engineering.
The studio takes web, mobile, SaaS, and custom software with the same standard: a senior person can explain the architecture, and the thing we demo is the thing we deploy.
FAQ
Web applications, mobile apps, SaaS products, custom software, and the interfaces for them. AI is a core part of the studio: agents, multi-agent workflows, voice agents, LLM apps with retrieval, automation, and private deployments.
No. Plenty of the work is a product build that never needs a model. We bring AI in when it makes the product better, and we say so when it does not.
With a conversation. Discovery gives both sides a scope. If it is a fit, a prototype follows in days, then a production build. Email info@myceliumtech.co and say what you want in production.
A narrow, working slice of the riskiest part. A call flow you can dial, a screen you can use, or a retrieval loop over a sample of your documents. It is there so you can decide, and it is built to be extended.
Yes. We deploy on private AWS, and we can use local models when data cannot leave your environment. The accounts stay yours.
A real-time phone agent aimed at low latency, typically on LiveKit and Twilio. We design the call flow, the handoff to a person, and the record you need in order to review what was said. Restaurants, law firms, and taxi services are the patterns we know well.
You do. The code, the designs, and the cloud accounts we set up in your name stay with you.
From the scope, not from a menu. Discovery is where a defined piece of work gets a price. We would rather decline a poor fit than invent a package.
Contact
A short note is enough. We read it, and we reply with whether we are the right studio and what a first prototype would cover.