Hi, I'm Arjun!

I build and train real-world AI systems by blending first-principles curiosity with a slightly obsessive love for clean, explainable code. This corner of the internet is where I talk about how I work, what I'm learning, and the kind of problems I help organizations untangle long after the first prototype ships. I'm an ML Scientist who loves dealing with a challenge head-on, identifying the solution, and getting hands-on with the training loop and the data quirks that shape it.

For HR Recruiters

When an HR partner asks what it's like to work with me, I start with the unglamorous bit: I finish what I promise without needing reminders. Years in computer vision and applied AI taught me to begin with the market need, build the right scaffolding around data and evaluation, and stay with the feature until its signals are stable in production. I like compact, high-trust teams where people own their work, communicate honestly, and obsess over the craft because the stakes are real and that's where the team-player version of me actually shines. Give me a focused pod and I'll happily set the cadence, keep priorities honest, and still leave space for curiosity that keeps the exploration active. I fuss over documentation and pipeline hygiene because future-us shouldn't reverse-engineer intent, and my work experience page captures those stories. When goals feel fuzzy, I pull the right stakeholders together, review the direction, and leave with a plan the team can trust across org boundaries.

For Techies

Most of my time has gone into computer vision, generative modeling, and the tooling around them—spanning supervised pipelines, self-supervised pretraining, synthetic data experiments, and the production frameworks that make all of that debuggable. I'm happiest when I can toggle between research-y exploration and the engineering that gets models into production, keeping experiments traceable and decisions transparent. Breadth matters to me: I'd rather understand how the whole stack clicks together and then deepen where the bottleneck appears than live forever in a single niche. Shipping taught me to treat observability, explainability, and client-safe configs as part of the model itself, which is why I built our production codebase from the ground up so new models, retirements, and bugfixes all share the same spine. I keep up with new papers, enjoy picking them apart with teammates or juniors, and blog the takeaways so those write-ups become the notebooks future-me can revisit.

What I'm exploring lately

Outside the day job, I've been playing with ideas from market structure and scalping strategies, tinkering with habit-building systems, and getting more comfortable with infrastructure topics like Kubernetes and microservices so I can collaborate better across the stack, and it's the kind of exploration I end up sharing with friends and curious strangers alike. All of these threads point to the same obsession: designing systems—technical or personal—that stay reliable under pressure and continue compounding over time. The next horizon for me is general-purpose and multimodal AI, where my computer vision instincts can anchor a group that's stitching together text, vision, and other structured signals for real products. I'd love to lead an R&D crew that spends half our time exploring and half exploiting, keeping growth for me and the company tightly aligned while we glue every layer of the stack together.