CSE student who profiles GPUs before breakfast — tracing CNN bottlenecks on NVIDIA H200 clusters, fine-tuning LLMs with LoRA, and shipping front-end work on the side. Currently looking for the next hard problem.
A running log of internships, infra work, and open-source contributions — most recent first.
A curated set of builds — from GPU infrastructure to IoT hardware and open-source contributions.
Profiled a custom CNN on NVIDIA H200 with Nsight Systems, then fine-tuned open-source LLMs with LoRA/PEFT and scaled throughput with FlashAttention, AMP, and Tensor Parallelism.
View on LinkedInAn AI-based system designed to detect and prevent proxy attendance manipulation in academic settings.
View on GitHubEnergy-efficient smart streetlight system on Arduino, using light, motion, sound, and vibration sensors for adaptive lighting and accident detection.
View on GitHubMobile-first UI in Figma helping villagers check water availability, request emergency tankers, track usage, and get arrival notifications.
View on GitHubAn ongoing log of Data Structures & Algorithms problem-solving in Java — daily practice and pattern-building.
View on GitHubOpen-source contribution to a production-ready toolkit for running AI models locally, on-device.
View on GitHub