There is a special sort of optimism in the phrase “legacy system.” It suggests an inheritance. Often, what has been inherited is several million lines of code, a shrinking guild of people who understand them, and a manual last revised when flip phones were enviable. The software still moves money, assigns benefits, calculates prices or enforces policy. It also remembers thousands of decisions that the organization has forgotten it made.
Miten Marfatia built a career around that inconvenient memory. He founded EvolveWare in Santa Clara in 2001 with a precise ambition: automate the documentation, analysis, maintenance and modernization of software applications. The verb that matters comes first. Before old code can be replaced, moved or divided into microservices, somebody has to understand what it does.
This is not the cinematic side of technology. There are no cheering crowds when a dependency map is accurate. Yet the boring parts of software have a sly habit of becoming the important parts. A payroll rule nested in COBOL can be more consequential than a glossy new interface. Remove it by accident and modernity suddenly feels overrated.
An engineer learns to price risk
Marfatia's path to software archaeology began with electrical engineering. He earned his undergraduate degree at the University of Bombay, then went to the University of Wisconsin-Madison, where he completed a master's in electrical engineering and an MBA focused on finance, investment and banking. The pairing now reads like an early draft of his operating method. Engineering asks whether a system works. Finance asks what failure costs, how uncertainty should be priced, and what evidence earns the next dollar.
His first documented company chapter was Silicon Electronics, which distributed advanced products to India's emerging computer-manufacturing industry. He led it from the mid-1980s. He later co-founded Perisol Technology, a supplier of storage solutions to corporations and government agencies. Hardware distribution, storage systems, enterprise software: each job sat close to infrastructure that customers could not casually switch off.
When EvolveWare arrived in 2001, the dot-com mood favored the new. Marfatia chose the accumulated old. The company's early products, Legacy Modernizer and Legacy Maintainer, appeared between 2005 and 2008. Their names are almost comically literal, like two dependable tradespeople arriving with labeled toolboxes. In 2015, their capabilities came together in Intellisys, a workflow-based platform designed to follow applications from maintenance through modernization.
The patent is a map of the obsession
The clearest record of Marfatia's technical preoccupation sits in the patent archive. He and co-inventor Ajay M. Rambhia are named on EvolveWare patents covering a pattern-abstraction engine, a contextual-pattern decoder, knowledge extraction and transformation, and an apparatus for migrating software code between platforms. One family of inventions describes taking source code, abstracting it into a standard stream, converting it into XML, then using the result for documentation, migration, impact analysis or security analysis.
In plain English: make the code explain itself in a neutral form. The idea predates the current enthusiasm for asking language models to read a repository and write a summary. EvolveWare describes its knowledge base as having evolved over more than 20 years with artificial intelligence implemented through machine-learning algorithms. The vocabulary has changed. The underlying problem has not.
Code is unusually bad at introducing itself. Its official purpose may be described in a binder; its actual purpose lives in branches, patches and exceptions. A rule written for a policy change in 1998 can survive three reorganizations and everyone who remembers the meeting. Extracting business rules is therefore less like translation and more like institutional fieldwork.
The patent chronology shows that the work was not a weekend fascination. Applications in the family reach back to the mid-2000s; grants followed across 2010, 2011 and 2013. The inventions separate recognition into parts: identify patterns, decode their context, collect the knowledge, then transform or migrate it. That modularity mirrors the management philosophy Marfatia would later describe in his articles. Large uncertainty becomes tractable when it is divided into named stages with outputs people can inspect.
There is also an understated ambition in the phrase “neutral form.” Programming languages come and go, but a tax calculation or eligibility test belongs to the organization, not to COBOL, Java or whichever platform happens to host it. Separate the rule from the syntax and the organization recovers a choice. It can document the rule, examine the consequences of changing it, or carry it into a different architecture. The code stops being the only place where the business can remember itself.
A pilot earns the sequel
Marfatia's public writing is more procedural than prophetic. That is its charm. In a 2023 article, he described a large state agency that needed business rules extracted from a multimillion-line application. Earlier outside efforts had not delivered what the agency required. Instead of repeating the grand attempt, a stakeholder chose a module for a pilot and divided it into smaller code segments.
The pilot lasted more than four months. Intermediate results were reviewed. The platform was customized. Processes were tested, changed and written down. Only after an end-to-end run produced rules in the required format did the agency specify the platform and process in later bids. Marfatia reported that the broader effort extracted rules from more than 20 million lines of code over more than two years.
The useful lesson is not the giant number. It is the small beginning. Marfatia's phased method asks every stage to produce an artifact that can be examined: an assessment, a map, a rule, a tested module. Budgets attach to success points. The modernization program becomes a series of auditable promises instead of one magnificent promise due several years from now.
That caution does not mean timidity. It is a method for changing consequential systems without pretending their complexity can be wished away. In his writing for government technology leaders, Marfatia has tied modernization to digital access, cybersecurity and the scarcity of people able to maintain existing portfolios. His point is practical: old systems become public-service problems when citizens expect remote, responsive access and the technical workforce is thinning.
His 2022 Techstrong TV appearance placed that argument in conversation rather than a product sheet. Across 18 minutes with host Michael Vizard, the topic was why application modernization deserved more attention. The frame matters. For organizations, modernization competes with every visible new initiative while its benefits often arrive as absences: the outage that did not occur, the change that did not break a dependency, the specialist whose retirement did not take the only map of the system with them.
The translator in the executive suite
Marfatia's role requires translation in both directions. Engineers need a faithful account of what an application does. Executives need a faithful account of what changing it will cost and why it matters. He has written that technical departments may prioritize productivity or reducing dependence on legacy code while the wider organization measures customer experience and cloud readiness. If the two groups use different scoreboards, even a sensible project can remain unfunded.
His finance training feels present here. A pilot is not merely a technical experiment; it is a financial instrument. It narrows uncertainty. Intermediate deliverables let leaders see what has been purchased. A repeatable process turns one success into a credible forecast for the next module. The best case for modernization may not be that the old code is embarrassing. It may be that its risk can finally be described.
EvolveWare's history carries the slow rhythm of enterprise change. The company says its technology received five US patents and one Australian patent. IT Week placed EvolveWare among its Top 50 Technology Innovators in 2005 and 2006, and the company's account says Service Provider Weekly named Marfatia its Person of the Year in 2006. Two decades later, EvolveWare says it was listed as a representative vendor in a 2026 Gartner market guide for AI-augmented code-modernization tools.
Patience as product strategy
Being early is flattering in retrospect and expensive in real time. EvolveWare began pursuing automated code understanding before machine learning became a boardroom phrase. Marfatia stayed with the problem through several generations of industry language: migration, transformation, cloud, microservices, AI. Each new label changed what customers wanted to buy, but not the awkward first task. Somebody still had to find out what the existing system knew.
His story offers a useful correction to the cult of the clean slate. Complex organizations rarely receive one. They inherit. They patch. They acquire. Their systems become geological, with eras of policy and preference pressed together. The responsible modernizer does not dynamite the formation and hope the fossils reassemble in the cloud.
Marfatia's alternative is almost domestic: inventory what you have, label what matters, repair one room, inspect the work, then continue. It is less thrilling than demolition and far more compatible with a government agency that must open on Monday.
After 25 years at EvolveWare, his aspiration remains visible in the company's original premise: a single platform that can help manage applications across their lives, from maintenance to modernization. The future he argues for is not one where old code disappears overnight. It is one where organizations can understand their past well enough to change it without losing themselves.