"Fundamentally, hum curiosity driven log hain"
What Drives Saurabh Karn's Journey in Legal AI with OpenNyAI and Jugalbandi
A mid-week afternoon, over a random online check-in, I asked my colleague Saurabh Karn where the seeds for his work with the OpenNyAI mission really get sown. I was curious because how many times do we pause to reflect on the moments that have led to the work we are doing now, the chance conversations, and the innate curiosities that drive each of us to keep learning and growing.
This is a snippet from my chat with him.
For the uninitiated, the idea of AI’s application in law and justice isn’t new. It stems from the old ideas that have already existed around legal automation, i.e. that repetitive legal processes can be automated. The industrial revolution, really.
It’s a subject that clearly excites my colleague, Saurabh. One can tell by how he moves closer to the laptop to tell me more. He’s recently been featured in a news article on his work with AI in Bangalore, and the shy grin when you bring it up is infectious.
He continues.
Machine learning has recently advanced significantly due to large datasets, powerful computing hardware, and open-source software becoming available. Geoffrey Hinton’s contributions, including the backpropagation algorithm in the 1980s and the introduction of transformers in 2017, revolutionised neural networks. These developments have enabled the training of large neural networks on massive datasets, leading to solving previously impossible problems.
From Curiosity to Collaboration
Where does that connect with the legal AI work he and his team have been hammering away at as part of the OpenNyAI mission, I ask.
In the case of legal AI, he says a lot of work was done by academia. In India, various technical institutions also conducted research and development, creating models and evaluation criteria. Something that we saw happening largely in a commercial setting previously.
Spend just minutes with him, and you will know that curiosity is Saurabh’s driving force. As a student at his engineering college, he bunked classes to attend classes on other subjects. He tells me of the time he decided to work in the chemistry lab of the Energy Department, tinkering with polyaniline— a chemical material to make many things like superconductors, actuators etc. It’s a toxic material, he adds. So he worked on making it green. And was successful. “Kaafi fun tha (Was a lot of fun)” is how he describes it. That’s how Saurabh describes most things, actually. The happy accident (think cloudy chem lab explosions et al.) that led to its making is one that he shares with barely disguised glee.
As he leans back into his chair, he contemplates, “Duniya kaise kaam karti hai, pata karna hai (How does the world function, that’s what I want to find out)”. If he got it tattooed somewhere, I wouldn’t be surprised, given the intrepid discoverer he is. And once something catches his fancy, he really gets into it. From AI to well, cycling. During the lockdown, he picked up cycling as a hobby— ran with it more likely, because just within the month, he’d covered 500 km in Delhi.
An engineer by training but a teacher at heart (he is a TFI-ite* through and through), it was while working on the National Education Policy as part of Central Square Foundation (CSF) that he started looking at the law. Policymakers need to look at the law, and that’s what opened this field to him, he mentions. His biggest inspiration to work in law and justice, though, he says, is the movie Legally Blonde. I guffaw at this. So he explains.
The movie, he says, showed him that it doesn’t matter where one came from. “See, she might have gotten into it for the wrong reasons, but she wins her final court case because of her knowledge of chemicals used in a hair perm. Anyone can do anything using logic”, he carries on enthusiastically. At one point, he also attempted the LLB exam at Delhi University while he continued working at CSF. He also looked at getting a Juris Doctorate abroad and was especially intrigued by the work being done at CODEX, Stanford at the time— a plugin that pulled on many housing laws to create a physical house model for you that fit the legal criterion. The law made ready to use for a citizen, or as he started mulling - is the law a high-frequency use case for citizens?!
Exploring Challenges and Discovering Solutions
With a colleague who was just out of Harvard, he started exploring the idea of working on legal datasets. That went nowhere. They had no legal expertise. Around this time, in 2019, he happened to chance upon a webinar on unlocking the full potential of legal data hosted by Sachin Malhan from the Agami team. Vivek Raghavan, of the UIDAI team that gave the country Aadhar, was part of that panel.
A few months on, Saurabh sent Sachin a message on Linkedin:
“I am keen on starting something in the field of legal tech, potentially in the field with a focus on NLP. I had a chance to see the Learning from Startups video on Agami's youtube channel, and one thing that remained with me is something that you had said - For the next 15-20 years there will be enough room for players in this space. I am keen on getting into the field albeit my understanding of Law is limited. I was wondering if you would be open to providing some guidance on how to approach this.”
Cut to a few months later when Saurabh actually ended up joining the Agami team as part of the open data mission in law and justice. A collaborative effort was underway to build Justice Hub as a repository of clean and usable legal datasets. As part of it, the team launched ‘Summer of Data’ - a 4-week program for crowdsourcing data by working with 65 law students. He calls the experience “savage”.
He says law students are no easy group to handle, but working with them brought him new learnings, especially through the expert talks he curated as part of the program. They opened Pandora’s box to much more. He goes on, “Lack of judgment-related data, private publishers having copyright over the formatting of judgements, pendency, hiring of judges— all these were recurrent themes. Even to show that there is no data, needed data— and working with these students allowed us to do that”. Three years since, the high demand for that work has proved to him the need and also what shared success with a community looks like. Many of the same students have continued working on other things with him. Speaking to students who he says don’t have the baggage of the past and can ask obvious questions is what he finds most fun - “Seasoned lawyers will not ask the questions on why some things exist or don’t like the way they do.”
Collaborative Making and Future Curiosities
It’s in these things and a way of working that he finds he operates best. Since 2022, as part of OpenNyAI, where he works with multiple experts and entrepreneurs in the field, he is quick to point out that what really works for him is the peer-to-peer work culture where they are all organised around the idea, rather than who is the organiser. “We call out each other’s silly ideas but also listen. These are, to me, the core ingredients of a collaboration that can sustain. We’re all makers at heart.”
And they have been up to some serious making. Last December, Jugalbandi was launched publicly at the Agami Summit. That it has exploded since would be an understatement. Satya Nadella spoke of it during his visit to India in January 2023, and in May, this was featured at the mecca for developers globally - Microsoft Build 2023. An internal joke for team Agami is the jugalbandi between Saurabh and Smita, both of whom get a somewhat crazy glint in their eye every time AI and its possible use cases come up in a conversation—mirrored by Sachin (the three of them work the most closely on this). However, he does a better job of playing it cool.
In what Saurabh is curious about in the weeks ahead, his answer is pretty straightforward, “First, how can we make citizen’s lives easier using a Jugalbandi-like stack which is powering multiple use cases across organisations? Secondly, I am very curious about the role Large Language Models (LLMs) will play in legal product designs. Thirdly, how regulatory and legislative orders can be more efficiently disseminated. We are working on a few experimental reference solutions of AI in the form of Jugalbandi powered stack - JIVA, Jagrit etc.”
Suffice it to say that there is a lot in the works that is keeping him and his team busy and excited. Because “Fundamentally hum curiosity driven log hain”, Saurabh concludes. For now.
To connect with Saurabh or learn more about his work with OpenNyAI, write to him at saurabh@agami.in.