Software studio

Software, shipped with senior engineers.AI, when it belongs in the product.

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.

  • OpenAI
  • Anthropic
  • Open-source LLMs
  • LiveKit
  • Twilio
  • AWS
  • Supabase
  • Vercel
  • Python
  • TypeScript
  • React
  • Next.js

Services

A product studio with a deeper bench in AI.

Two practices, one team. Product engineering for the thing people touch. AI engineering for the part that listens, decides, or answers.

AI engineering

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.

02

Voice AI agents

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

LLM applications and RAG

Products that answer from your documents. Retrieval, citations, and the decision to decline an unsupported answer are part of the build.

04

AI on your own infrastructure

Private AWS deployments, including local models, for teams whose data cannot leave their network.

05

Business workflow automation

The repetitive movement of data between tools, replaced with a flow your staff can see, pause, and trust.

06

AI product MVPs

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

Web applications

Product sites, customer portals, and internal web apps with the performance, accessibility, and structure a team can keep owning.

02

Mobile apps

iOS and Android products with deliberate flows, a release path, and an interface that matches the web product beside them.

03

SaaS products

Multi-tenant products with accounts, roles, billing, and the admin tools a real customer base requires.

04

Custom software

Systems shaped around one operation: the handoffs, permissions, and data your team already lives with.

05

UI/UX

Interface design done with the people who will build it, so the screens you approve are the screens that ship.

Use cases

Where the work shows up.

The same studio builds the product and the system behind a phone call, a document, or a private deployment.

01

Restaurants

A phone agent that takes reservations and orders, and hands the messy exception to a person.

02

Law firms

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.

03

Taxi and dispatch

Call handling that books a trip, updates it, and passes the details to dispatch.

04

Product teams

Web and mobile products, from a first release to the version a customer can pay for.

05

Operations teams

Workflow automation across the systems people already have open all day.

06

Private environments

Assistants and models running in your own AWS account when the data has to stay there.

How we work

From a conversation to a system you own.

  1. 01

    Discovery

    We learn the user, the constraint, and what done means. You leave with a scope you can accept or refuse.

  2. 02

    Prototype in days

    A working slice you can click, call, or hand to a colleague. Enough to make a decision, and built so it can continue.

  3. 03

    Production build

    Design, engineering, and the parts that keep it alive: auth, data, logging, and a deploy you own.

  4. 04

    Support

    We stay after launch for fixes and the next increment, with the context of the people who built it.

Capabilities

What a build can look like.

A system is easier to picture than a slogan. These are the kinds of work we design and ship.

Voice

Restaurant phone desk

  • A voice agent on a real phone number
  • Reservations, hours, and a short set of order questions
  • A clean handoff when the caller needs a person

Voice

Firm intake line

  • An after-hours flow that collects the caller and the matter
  • A structured summary waiting in the morning
  • No performance of legal advice

LLM

Document assistant

  • Answers drawn from a defined set of files
  • A citation for where the answer came from
  • A refusal when the files do not support it

Product

SaaS first release

  • Accounts, roles, and billing
  • One core workflow, finished
  • A product a customer can pay for

Product and voice

Dispatch with a phone agent

  • A web console for trips and drivers
  • A phone agent in front of it for bookings and status
  • The exception path lands with a dispatcher

Infrastructure

Private model deployment

  • An assistant running in your AWS account
  • Access limited to your network
  • A path to a local model when that is the requirement

Capability tiles, not client stories. Named projects are on the work page.

Selected work

A few of the systems in production.

Lead projects from the studio. The full list, including work still in progress, is on the portfolio page.

Stack

Tools we use when they fit.

A palette, not a promise that every project uses all of them.

OpenAI and Anthropic

Hosted models when a frontier model fits the product and the data policy allows it.

Open-source LLMs

When we need to choose the model, or run it ourselves.

LiveKit and Twilio

The real-time path for phone agents.

AWS

Private deployments, and the rest of the infrastructure around them.

Supabase and Vercel

Application data and hosting when they are the right fit.

Python and TypeScript

The languages the studio builds in.

React and Next.js

The web interface layer.

Studio

Led by someone who still builds.

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

Before the first call.

Contact

Tell us what you want in production.

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.