From: ThanhVu (Vu) Nguyen
3 views
Skip to first unread message
ThanhVu Nguyen
Apr 25, 2026, 10:48:59 AM (yesterday) Apr 25
to deepak-kapur-memorial
= ThanhVu (Vu) Nguyen
== Former PhD Student of Deepak's · George Mason University
It was a great shock when Roli told me about Deepak's passing on Saturday. We were in the middle of an NSF proposal submission, and Deepak had just sent me suggestions on Wednesday. Despite his deteriorating health, he was still very active in research and eager to learn new things, and his thinking was very sharp in his last days as it was when he was in his prime.
In the beginning of my PhD study at UNM, I was working on program repair using genetic programming, and we were quite successful in fixing real bugs. However, Deepak was skeptical about the lack of formal specification. He kept asking about the meaning of "correct" and "bug", which I could not provide, and this led him to question the work: "how can you fix what you don't understand?" Deepak transitioned me to his "rigorous" camp, and my PhD thesis is about dynamic invariant generation and more generally formal program analysis, which forms the basis of all my work since then.
Despite being a respected senior researcher, Deepak could be quite amusing and fun to be around. I always called him "Professor" (just Professor and not Professor Kapur) from the very beginning of my PhD. He found that amusing and started to call me "Professor" as well, which is how we addressed each other until I actually became a professor, at which point he started using my name. He also liked free food, and when he saw announcements from UNM events with food, he would send me to go and get some for him. I remember that once the Daily Lobo had an ad for a restaurant offering a heavy discount — \$1 for a taco/sandwich — and I grabbed 20-something newspapers from the Daily Lobo office, and we went there for lunch for almost a month. Clearly mischievous, but also a lot of fun.
When my first paper with him received the Distinguished Paper award at ICSE — the top conference in software engineering — he said "that tells him a lot about the quality of SE research." But that is how he was: brutally honest and never mincing his words, which of course could easily offend people. But if people could get past that, as I and many others who knew him well can attest, they would find that Deepak cared a lot for his colleagues, students, and especially about CS education and research — perhaps a bit too much, such that he got frustrated when he saw "junk" research or students lacking foundational knowledge. I should note that many CS faculty today feel exactly the same way about the state of education in the age of AI and LLMs. The difference is that most of us are not as blunt as Deepak and therefore have much better RateMyProfessor ratings.
Deepak is the type of person that I simply do not believe could "retire" and would think about research until his last day. I have been working on neural network verification (NNV) in recent years, and Deepak was quite interested in this line of research and tried to learn about it — at the age of 73, knowing he had just a couple of years left! In fact, just a few weeks ago, I was describing to him my new paper on generating proofs to certify results of NNV tools, and he immediately observed that the resulting proofs can be used to modify the original networks and make them much smaller. I was quite excited about this idea and decided to invite him to be a co-PI on an NSF proposal I am submitting. His last message to me (Wed 4/8/2026 8:57PM) was about research. I imagine that in those last few days, he was still thinking about how to make NNV proof checking more efficient.
== Former PhD Student of Deepak's · George Mason University
It was a great shock when Roli told me about Deepak's passing on Saturday. We were in the middle of an NSF proposal submission, and Deepak had just sent me suggestions on Wednesday. Despite his deteriorating health, he was still very active in research and eager to learn new things, and his thinking was very sharp in his last days as it was when he was in his prime.
In the beginning of my PhD study at UNM, I was working on program repair using genetic programming, and we were quite successful in fixing real bugs. However, Deepak was skeptical about the lack of formal specification. He kept asking about the meaning of "correct" and "bug", which I could not provide, and this led him to question the work: "how can you fix what you don't understand?" Deepak transitioned me to his "rigorous" camp, and my PhD thesis is about dynamic invariant generation and more generally formal program analysis, which forms the basis of all my work since then.
Despite being a respected senior researcher, Deepak could be quite amusing and fun to be around. I always called him "Professor" (just Professor and not Professor Kapur) from the very beginning of my PhD. He found that amusing and started to call me "Professor" as well, which is how we addressed each other until I actually became a professor, at which point he started using my name. He also liked free food, and when he saw announcements from UNM events with food, he would send me to go and get some for him. I remember that once the Daily Lobo had an ad for a restaurant offering a heavy discount — \$1 for a taco/sandwich — and I grabbed 20-something newspapers from the Daily Lobo office, and we went there for lunch for almost a month. Clearly mischievous, but also a lot of fun.
When my first paper with him received the Distinguished Paper award at ICSE — the top conference in software engineering — he said "that tells him a lot about the quality of SE research." But that is how he was: brutally honest and never mincing his words, which of course could easily offend people. But if people could get past that, as I and many others who knew him well can attest, they would find that Deepak cared a lot for his colleagues, students, and especially about CS education and research — perhaps a bit too much, such that he got frustrated when he saw "junk" research or students lacking foundational knowledge. I should note that many CS faculty today feel exactly the same way about the state of education in the age of AI and LLMs. The difference is that most of us are not as blunt as Deepak and therefore have much better RateMyProfessor ratings.
Deepak is the type of person that I simply do not believe could "retire" and would think about research until his last day. I have been working on neural network verification (NNV) in recent years, and Deepak was quite interested in this line of research and tried to learn about it — at the age of 73, knowing he had just a couple of years left! In fact, just a few weeks ago, I was describing to him my new paper on generating proofs to certify results of NNV tools, and he immediately observed that the resulting proofs can be used to modify the original networks and make them much smaller. I was quite excited about this idea and decided to invite him to be a co-PI on an NSF proposal I am submitting. His last message to me (Wed 4/8/2026 8:57PM) was about research. I imagine that in those last few days, he was still thinking about how to make NNV proof checking more efficient.
Reply all
Reply to author
Forward
0 new messages