Final-year BS-AI student crafting work that spans two layers most people keep separate — the operating system underneath, and the models running on top of it.
I'm an aspiring AI engineer and systems developer with a practical foundation that spans the operating system and the model running on it. Recent work includes a userspace daemon that reaches into the Linux scheduler directly, a CCTV violence-detection pipeline judged on real accuracy numbers, and a distributed data pipeline benchmarked against plain Pandas. I like problems that don't stay inside one layer of the stack.
C, Linux kernel APIs, /proc introspection, CPU scheduling & affinity.
TensorFlow/Keras, CNN/RNN/LSTM pipelines, transfer learning.
Dask Distributed, DAG-based execution, out-of-core processing.
Propositional & first-order logic, description logic, resolution proofs.
C · Linux Kernel API · /proc · dirent.h
A high-performance userspace daemon that tackles resource contention and micro-stuttering in Ubuntu's Completely Fair Scheduler, iterating through /proc for zero-overhead process introspection.
/proc parsingsched_setaffinitysetpriorityPython · TensorFlow/Keras · MobileNetV2
An end-to-end video classification pipeline judging CCTV footage across three classes: Normal, Violence, and Weaponized.
Python · Dask Distributed · Pandas · DAG
A decoupled master-worker parallel pipeline benchmarked against sequential Pandas on a synthetic 3,000,000-record dataset.
Propositional & First-Order Logic · Description Logic · Resolution Proofs
Real-world decision scenarios formalized across five knowledge representation paradigms.