Meet the Writer: Hacker Noon's Contributor Lokesh Prakash Manohar, AI Product Lead

\ Welcome to HackerNoon’s Meet the Writer Interview series, where we learn a bit more about the contributors that have written some of our favorite stories . So let’s start! Tell us a bit about yourself. For example, name, profession, and personal interests. I'm Lokesh. I’m an AI Product Lead at LatentView Analytics, where I work on enterprise AI applications, including RAG systems, knowledge graphs, and agent-based workflows. Over the years, I’ve had the opportunity to build solutions for teams at Google, YouTube, and Microsoft. Before LatentView, I spent time at a couple of start-ups, PayPal and LendingTree in data science and product roles. Outside of work, I co-lead the Research Triangle chapter of The AI Collective, contribute to open-source projects when I can, and follow the markets more closely than I probably should. \ Interesting! What was your latest Hackernoon Top story about? It started as a simple curiosity. When Google redesigned the Workspace icons, most people were debating whether the new look was better or worse. I found myself wondering how vision-language models would react to the change. Since computer-use agents often identify applications from screenshots, I wanted to know whether the new gradient-based icons are actually easier for vision encoders to tell apart than the older flat-color versions. I built a small, reproducible experiment to test it across several encoder families. The results were interesting, but so were the pitfalls. One thing I ran into was that many text-aligned models have the old icons baked into their training data, which can skew comparisons if you're not careful. The write-up ended up being as much about methodology as it was about the results. \ Do you usually write on similar topics? If not, what do you usually write about? Most of my writing falls into two categories. The first comes directly from problems I've encountered while building AI systems. If I run into something interesting in production, I'll usually spend time understanding it and then write about what I learned. That has led to articles on topics like prompt injection defenses, multi-hop reasoning in GraphRAG systems, and the hidden costs of migrating embedding models. The second category is more focused on strategy. I spend a lot of time thinking about where enterprise AI is heading and how organizations should respond. I've written about why general-purpose embedding models can limit retrieval quality, what it means to own your embedding layer, how to evaluate AI maturity inside large organizations, and why verifiability is becoming increasingly important as AI systems move into production. I try to keep those two areas connected. Strategic ideas are only useful if they're grounded in real-world experience. Likewise, technical experiments are most valuable when they help answer questions that matter to leaders making decisions about AI adoption and investment. The Workspace icon study followed the same pattern. It started as a question about a design change, but what interested me was the larger implication. As AI agents increasingly interact with software through visual interfaces, are those interfaces becoming easier for machines to understand? That's ultimately a strategic question, but one that required an empirical investigation. \ Great! What is your usual writing routine like (if you have one?) Usually it starts with something I can't stop thinking about. I'll have an idea, a question, or a claim I want to test, and before I write anything, I try to verify it myself. That might mean digging through data, building a quick prototype, or running a small experiment. The actual writing is the easy part. It usually gets done over a few evenings or weekends. The harder part is editing. My first drafts are always too long and full of caveats, so I spend a lot of time trimming them down and getting to the point faster. I also like to look at a few recent articles from wherever I'm planning to publish. Every platform has its own style and audience, and it's a good reality check for whether the piece actually belongs there. \ Being a writer in tech can be a challenge. It’s not often our main role, but an addition to another one. What is the biggest challenge you have when it comes to writing? Honestly, it's the gap between what I can verify and what would make a better story. Client work can't be discussed in detail, and the most interesting findings often need anonymizing to the point where they lose their punch. So I end up running independent experiments on public data to make the same points, which doubles the work. The other challenge is just time. A rigorous piece with real numbers takes weeks of evenings, and it's tempting to publish something thinner and faster. I've found the slower pieces are the only ones worth doing. \ What is the next thing you hope to achieve in your career? I'd like to keep moving toward the intersection of research and product: shipping AI systems that work in production while publishing what I learn along the way. More concretely, I want to grow my open-source work into tools other practitioners actually depend on, and keep contributing to the academic side through papers and reviewing. The longer-term goal is to be someone whose work has helped make AI systems in the enterprise more verifiable and less of a black box. \ Wow, that’s admirable. Now, something more casual: What is your guilty pleasure of choice? Watching options flow and market screens when I should be doing something else. I run a small systematic trading setup as a hobby, and I check it more often than the system actually requires. It's research, I tell myself. \ Do you have a non-tech-related hobby? If yes, what is it? Long strolls and hiking, mostly. There are plenty of greenways and trails near where I live, so it's easy to get out regularly. I also enjoy reading about culture and history. \ What can the Hacker Noon community expect to read from you next? Two pieces are in the works. One is about knowledge-graph-based RAG systems and why they're so hard to debug when they give wrong answers. I've been building something in that space and will write about it once it's ready. The other is about verifiability in enterprise AI, which I think is one of the most underrated factors in whether AI projects actually ship. That one is more of an argument piece, and I want to back it up properly before publishing. \ What’s your opinion on HackerNoon as a platform for writers? What I appreciate most is that it rewards substance over credentials. You don't need to be at a famous lab for a well-executed piece to find readers. The editorial process is straightforward, and the audience is genuinely technical, which keeps you honest. If you publish something hand-wavy, someone in the comments will notice. That's a feature. For practitioners who write on the side, it's one of the lower-friction ways to put real work in front of people who can evaluate it. \ Thanks for taking time to join our “ Meet the writer ” series. It was a pleasure. Do you have any closing words? Thanks for having me. If there's one thing I'd say to other practitioners who are thinking about writing: the experiment you ran to settle an argument with yourself is probably more interesting to readers than you think. Write it up, show your numbers, and be upfront about what didn't work. That's the whole formula. Check out Lokesh Prakash Manohar’s HackerNoon profile here, and read more of his amazing stories! https://hackernoon.com/u/lokesh-ai-pm