MIRA · Ionate

The whole platform,
one conversation.

Ionate's platform can untangle and transform 40-year-old legacy code. I designed the self-serve, chat-based experience that lets a customer harness all of it by just talking, named it MIRA, and wrote its story.

Role
Self-serve conversational experience, naming and backronym, narrative and launch deck. For Ionate.
Built
The self-serve, chat-based experience that puts Ionate's modernization platform in a customer's hands
Context
Ionate, enterprise legacy modernization (banking, mainframe COBOL)

A note. MIRA is a live Ionate product with public demos. Everything shown here uses mock data, and it is shared with Ionate's blessing.

See it

MIRA, in action.

A developer asks MIRA to modernize a wire-transfer module. Ionate's platform maps the ecosystem, extracts the buried business rules, converts the code, and runs the tests, all reached through one self-serve conversation. Mock data.

The problem

Everyone tried Gen AI. It didn't work.

Ask any modernization team if they use Gen AI and the answer is yes. Ask if it works and the answer is a wince. A generic LLM sees a snippet, not a 40-year interconnected ecosystem, so it invents dependencies, forgets what a program does across its callers, and cannot actually convert, test, or ship anything.

No context
It sees code snippets, not the ecosystem they live in.
Hallucinations
It invents dependencies, business rules, and transform paths.
No memory
It forgets what a program does across 47 callers.
No execution
It suggests, but cannot convert, test, or deploy.

"It gives me plausible Java. It's wrong in ways I won't discover until production."

The reframe

Ionate solved the hard part. It was locked away.

Ionate's platform genuinely understands legacy code, how it connects and how to transform it safely. Real technology, years in the making. But wielding it took experts and a services engagement. A customer could not simply pick it up and use it.

That was the design problem I owned: not the modernization, but the access. Make that power self-serve, so a customer reaches the whole platform without a specialist on the call.

The power was there. It needed a front door.

The design

One teammate over three proven engines.

Ionate had spent years building three engines: discovery, transformation, and delivery. Powerful, and separate, and expert-operated. I designed a single self-serve, conversational experience that puts all three within a customer's reach, so she stops running tools and starts talking to one teammate who runs them for her.

I named her MIRA, gave the name its backronym, and built the story that carries the product, from the cold open to the close.

SOTERIA
Discovery
"Here's the lay of the land."
APPDATE
Transformation
"I'll convert this for you."
KÍRKĒ
Delivery
"Here's your deployment."

MIRA  /  Modernization · Intelligence · Reasoning · Agent

The user

Meet Sarah.

A senior developer at a legacy bank, handed the wire-transfer module: 340,000 lines of COBOL written between 1987 and 2003. The original authors have retired. Documentation is sparse. The code moves $2.3 billion in daily transactions. The story is built around her, because the product only matters if it works for her.

127 interconnected programs
23 external system interfaces
Unknown business rules embedded in code
Zero tolerance for production failures

Estimated timeline, the traditional way: 18 months.

The difference

She knows. She doesn't guess.

The platform reads the actual code. My job was to make that trustworthy in the conversation, so every answer comes back with the line numbers it was drawn from, and a customer can verify instead of taking it on faith. Ask a generic LLM about a fee program and it guesses from "typical banking patterns." Ask MIRA, and the answer cites its source.

Ask MIRA — business rules
What business rules are embedded in FEE-CALC-007?
FEE-CALC-007 contains 7 distinct fee rules:
1Domestic < $10K → flat $15
2Domestic $10K–$100K → 0.15%
4International SWIFT → $45 + 0.20%
6VIP accounts (tier 3+) → 50% discount
These rules are undocumented. Extracted from conditional logic at lines 234–892.

"MIRA doesn't guess. She knows."

Trust

Your code never leaves.

For a bank, that promise is the whole sale. MIRA runs in a secure, isolated container inside the customer's own infrastructure, so trust is a property of the architecture, not a line in a contract.

Air-gapped
Runs with no internet required. No source ever sent to external APIs.
SOC 2 compliant
Enterprise security standards, with full audit trails.
Your model, your rules
Bring your own model. No training on your proprietary logic.
Outcome

Eighteen months, to six weeks.

In the wire-transfer demo, an 18-month migration becomes a 6-week conversion. That power is Ionate's platform; my part was the self-serve experience that puts it within a customer's reach, by just talking. MIRA is live at Ionate, and the story I built is how it introduces itself to the world.

Next case Intuitive Surgical·The reorder signal, in one glance