Inference Engineer based in New York.

I'm Prem — an inference engineer focused on low-latency LLM serving, multi-model routing, and cost-efficient production AI infrastructure. I've deployed scalable AI systems on GCP and AWS with a hard bias toward latency, reliability, and the cost-latency-accuracy tradeoff — not whichever framework is trending this quarter.

Right now I'm finishing my M.S. in Artificial Intelligence at the Rochester Institute of Technology and shipping Emotion Engine — a research project showing that emotion-like internal states (fear, grief, suspicion) emerge from a 72-feature predictive LSTM with no hardcoded rules (p = 3.3e-113 across 205K agent-step records). Before RIT, I built a production LLM routing layer at Concentrix + Webhelp — 3+ foundation models, 1.5s end-to-end, 18% cost reduction, 42% fewer incidents.

The throughline of everything on this site: “Simple systems scale better than clever ones.”

Education

• M.S. in Artificial Intelligence — Rochester Institute of Technology, Aug 2024 – May 2026
• B.Tech in Computer Science and Engineering — National Institute of Technology, Silchar, Aug 2020 – May 2024

Experience

• Generative AI Engineer, Concentrix + Webhelp — Feb 2024 – Jul 2024
• Data Science Intern, AlphaBits Technologies — Aug 2023 – Jan 2024
• ML Engineer Intern, iNeuron.ai — Jun 2023 – Aug 2023

Publications

• Lightweight Channel Attention for Efficient CNNs — Designed and evaluated a lightweight channel attention module (LCA) achieving competitive accuracy with negligible parameter and latency overhead on ResNet-18 and MobileNetV2.

Certifications

• Machine Learning Specialization — Stanford Online · Coursera