About me

I care about building AI that people can actually trust.

My name is Lokesh, and I'm a researcher and engineer working on the messy, human side of artificial intelligence. Right now, I have finished my M.S. in Cybersecurity Analytics and Operations at Penn State, but my real focus is on a problem that sits somewhere between engineering and philosophy: how do we make AI systems safe enough, observable enough, and responsible enough that real people can rely on them?

Most of my work centers on understanding AI behavior—not just whether a model gets the right answer, but why it got there. I've spent a lot of time studying how large language models cave to pressure, repeat back what they think you want to hear, or generate confident reasoning that doesn't actually match their internal process. In my research on sycophancy and chain-of-thought faithfulness, I've seen models abandon correct answers when pushed, or rationalize biases they shouldn't have. That's a trust problem, and I think it's one of the most important problems in AI right now.

But I'm not just interested in breaking things in a lab. What motivates me is making AI observable and accountable in the real world—building systems where you can trace a decision back to its source, understand whether your agent is reasoning or just pattern-matching, and catch failure modes before they impact someone.

Right now, I'm living that challenge quite literally. My parents run an organic business, and I'm building an AI system to help them modernize it. I'm automating item procurement and delivery decisions—predicting inventory needs and optimizing logistics so they can focus on growing the business instead of managing spreadsheets. The part I'm most excited about is the customer side: evaluating a person's medical conditions to suggest organic items that fit their specific health needs, along with appropriate dosages. I want to be clear—this isn't a doctor, and it doesn't diagnose anything. It's a smart wellness assistant that offers personalized guidance, helping customers make informed choices about organic food based on their own health context. Because this affects my own family and our community, I can't afford to get the "responsible AI" part wrong. It has to be transparent, handle data carefully, and actually make people's lives easier.

This is the kind of business problem-to-AI solution translation I love. A lot of organizations know they want to use AI, but struggle to frame the problem in a way a model can actually solve. I enjoy that work—taking a real operational pain point, figuring out what data and architecture actually fit, and shipping something that doesn't just perform well on a benchmark but makes someone's day better. Whether it's detecting anomalous network traffic with 94% accuracy or helping a family business serve its customers more thoughtfully, I try to stay grounded in outcomes.

My path into this space started during my undergrad at Anna University, where I worked as an AI Researcher building neural network models to identify malicious network packets. I worked with Convolutional Neural Networks and Long Short-Term Memory (LSTM) architectures, and that experience became the cornerstone of my master's research on resource-efficient DDoS attack identification using BiLSTM and CNN ensemble models. I've also shipped production code at Imerit, where I built and managed user identification systems across their platform—making the onboarding process smoother and more secure for new employees.

On the technical side, I spend most of my time in Python, PyTorch, and the HuggingFace ecosystem, with backend work in FastAPI, Docker, and AWS.

If you're working on AI safety, building agents that need to be auditable, or figuring out how to deploy AI responsibly without sacrificing performance, I'd love to chat. I'm always open to conversations about making these systems work better for the people who actually use them

---

Let's talk: lokeshlks01@gmail.com

Other places: Blog | Chess

---

"We used to look up at the sky and wonder at our place in the stars. Now we just look down, and worry about our place in the dirt."

I'm still looking up.