Data Science and Engineering: Pushing Boundaries in Legal AI
Raviraj Prajapat
Data Scientist at Sirion
By Meghana Biwalkar
Raviraj Prajapat learned early in his childhood that curiosity could create new paths and opportunities for him. Years later, that habit still defines him.
As a data scientist at Sirion, he builds AI-led systems that read and reason through complex legal contracts, guided by the same curiosity that once pushed him to take things apart and see how they worked.
The Making of an Engineer
In his family home in Bikaner, near India’s western border, when something broke, whether a fan, a switch, or a bicycle, Raviraj was the one to fix it.
His father was a teacher, his grandfather a farmer, and new things were rare. Growing up, the rule at home was simple: repair, don’t replace.
What started as a necessity became a habit. At school, he tried to recreate physics experiments with whatever he could find handy. With each attempt, from simple experiments to complex ones, it became obvious to him that he learned more by taking things apart, questioning them, and putting them back together in his own way.
When it was time for college, he chose mechanical engineering as it felt like a natural extension of how he already thought—build, test, repeat. This instinct to reverse engineer shaped Raviraj’s mindset and ultimately drove him toward a career in engineering.
Tracing a Path from Workshop to Data
During his engineering days, Raviraj gravitated toward practical workshops as a way of learning. In one project, he and his college mates designed and built a car for a national competition—from scratch! They welded the chassis, fitted the gearbox, and tested it.
The project taught him a principle that he follows till date—scarcity can be a strength. “When you don’t have everything, you need, you learn to be inventive,” he recalls. “While building the car from nothing, we had to make do with what was available. That’s how I learned to value what really matters.”
Through college, Raviraj continued to focus on workshops and practical learning. A specific robotics project proved pivotal in his career decision. While working with his classmates to design a prosthetic device that could move using muscle sensors, he uncovered the power of coding. This was his introduction to programming. Watching a few lines of logic turn into physical motion changed that, as he realized programming was another way to engineer. These early hands-on projects significantly shaped Raviraj's understanding of programming's practical applications by demonstrating its direct connection to the physical world.
From Machines to Data Science
“I never stopped being a mechanical engineer,” he points out. “Only the tools changed.” After graduation, Raviraj worked in the automobile industry. He expected to be surrounded by machines; instead, he was surrounded by data. Every test run produced sheets of readings that helped his team cut short their testing time.
He learned to read data inputs like a story: when something went wrong, the numbers always told him first. Raviraj says, “Patterns in numbers started to reveal what might fail before it actually did. That curiosity pushed me beyond analysis and towards building models that could predict outcomes, not just explain them."
Through years of testing and data analysis, Raviraj stumbled upon an opportunity with Sirion. “I was intrigued when a friend mentioned Sirion, a company that uses AI to help global businesses manage enterprise contracts,” recalls Raviraj. “I had no idea what ‘enterprise AI’ meant, but I was curious enough to apply.”
Sirion was his first software job. Soon, “I realized contracts encode conditional logic, risk allocation, and negotiation history. It was just logic written in another language,” he says. “Every clause is a condition. ‘If this happens, then that’. The structure felt familiar; only the syntax was new.” But, unlike usual machine learning problems, false positives were penalized more in a contract than false negatives. "This understanding led us to rethink the evaluation metrics for contracts."
But the challenge was to teach the machine the complexity that was beyond simple logic: ambiguity, human judgment, and negotiation history.
Decoding a new language
To teach a machine a language, Raviraj first had to learn it himself. He spent hours with the leadership at Sirion, who guided him through the architecture of contracts. “Our CEO and co-founder, Ajay would explain why one word could change the meaning of an entire deal,” Prajapat recalls. “And I’d translate that logic into code.”
That process unlocked a new way of looking at contracts.
Raviraj started to map contracts in his head as if they were code: definitions connected to obligations. Rights pointed to conditions. Exceptions branched like nested loops. The realization felt familiar. The same way he once mapped circuits or engine systems. Contracts, too, had architecture.
“When I worked on these contracts, I always had this mind map,” he says. “I can read them as coding blocks. Each clause connects like a logical flow. That’s when I realized I am a translator, converting a contract into code.”
Building with experts
As Sirion grew, Raviraj knew data science alone wouldn’t be enough on its own. Contracts were social documents as much as technical ones, and training an algorithm to interpret them required the people who knew them best.
Enter the Legal Engineering team, a new discipline that pairs lawyers with AI specialists. “It's not just data annotation." He adds, "It's the act of converting complex legal interpretation into computational structures by unpacking the concepts and neatly weaving them into a legal knowledge graph. This bridges the gap in reasoning.”
So, when the lawyers in the team flagged poor accuracy or missed clauses, he dug deeper. “In most setups, you’d just note that ten things worked and three didn’t. But that’s not how AI learns. I’d ask why it failed. What was the scenario? What made it different?”
These conversations, often with Sirion’s legal engineering team, became the real testing ground. He’d take their feedback, probe further, and keep asking questions until he understood the root cause. “If I don’t know what failed, I don’t know what I’m fixing,” he says.
This back-and-forth shaped his philosophy, “If you don’t ask the right questions,” he says, “you will never get the right answers. Otherwise, you’ll have no clue what you are really creating a solution for.”
He also realized that most engineers rarely experience their own product as end users. “That’s the missing empathy,” he says. “Product teams do the research, but engineers need that connection too. When I started sitting with lawyers and watching how they used the system, I finally understood what we were building for.”
That shift changed everything. Instead of promising flashy outcomes, the team started experimenting fast, failing in hundreds of small ways, then scaling what worked. “We stopped guessing what users needed and started learning from their failures,” he says. “That’s how we built systems at Sirion—that actually hold up in the real world.”
Turning Contracts into Intelligence
Raviraj doesn’t measure success by the number of features. For him, success is when the system becomes invisible, when lawyers trust it enough to stop double-checking it.
“A lawyer shouldn’t spend hours hunting through PDFs,” he says. “They should spend time thinking. Our job is to make that possible.”
Today, he and the team at Sirion are working on the next-gen agentic CLM platform that helps identify risks, relationships, and intent in contracts—enabling companies make faster, more confident decisions. They evaluate risk, flag inconsistencies, and recommend negotiation positions, giving lawyers and procurement teams a head start on decisions that once took months.
“Contracts already contain intelligence, we’re just teaching machines to see it.”
- Raviraj Prajapat
Engineering with Restraint
Even at the edge of AI innovation, Raviraj’s instincts remain rooted in frugality. Build lean. Measure carefully. Iterate. “Throwing more power at a problem doesn’t make it smarter,” he says. “It just hides what’s broken.”
In a tech industry obsessed with spectacle, Raviraj’s work is defined by restraint. His ambition isn’t to build the flashiest model, it’s to build one people can trust.
His approach today follows the same logic. He builds smaller, explainable models instead of sprawling ones, designing systems that can show how they arrive at an answer. For him, trust matters more than speed. “If I can’t explain what the system did, I don’t trust it,” he says.
He doesn’t see contracts as paperwork but as living systems of cooperation. A record of promises, risk, and intent. “Every contract is a story of two sides trying to understand each other,” he says. “AI should help that story move faster and fairer.
Finding His Lane
At Sirion, Prajapat has found the rare overlap between precision engineering and abstract reasoning. Professionally, it’s the first place where his mechanical instincts and curiosity have met. Personally, it’s where he’s learned that innovation can be quiet.
Outside work, he paints and plays music. “Art teaches you to look for rhythm and proportion,” he says. “Code is no different.”
He draws a clear line between automation and augmentation. “Technology should reduce uncertainty,” he says. “Not responsibility.”
The boy from Bikaner who once welded car frames now welds together language and logic. The materials have changed, but the craft remains the same: build carefully, make it reliable, let it run.
Built with Shorthand