
RAVINDRAN S.
Systems Engineer · Edge AI Inference · Linux Kernel Interception
Computer Science undergraduate at Chennai Institute of Technology (9.24 CGPA) specializing in hardware-level systems programming, Edge AI model quantization, and deterministic offline runtimes.
Engineered sub-1 GFLOP object detection models for autonomous drones achieving 96% mAP with TensorRT INT8, built eBPF package verification systems in Rust, and contributed memory-safety patches to Amazon's AWS s2n-tls core cryptographic library.
CORE PRINCIPLES.
DETERMINISTIC LATENCY
Eliminating brittle user-space hooks. Using evdev exclusive grabbing (EVIOCGRAB) and /dev/uinput virtual HID synthesis to achieve deterministic <0.1ms latency.
HARDWARE BOUNDS FIRST
Engineering AI models to fit strict compute budgets. Calibrated TensorRT INT8 engines run sub-1 GFLOP vision at 62 FPS without thermal throttling.
CRYPTOGRAPHIC INTEGRITY
Security verified at the system level. Automated regression fuzzing, memory-safe C11 bounds in AWS s2n-tls, and rootless eBPF sandbox validation.
EDUCATION & RESEARCH.
Chennai Institute of Technology
B.E. Computer Science & Engineering (AI & ML)
Core Curriculum & Specialization: Data Structures & Algorithms, Computer Vision, Deep Learning, Operating Systems, Systems Programming, Distributed Systems, Computer Networks, and Edge AI Hardware.
TECHNICAL COMPETENCIES.
Problem Solving & Algorithms
LeetCode Knight (2085 peak contest rating) with rigorous discipline in dynamic programming, graph theory, and concurrent algorithmic pipelines.
Edge AI & Hardware Quantization
Engineered sub-1 GFLOP real-time drone vision models with TensorRT INT8 quantization running at 62 FPS on embedded hardware payloads.
Low-Level Systems & Security
Engineered package managers in Rust with Aya eBPF kernel monitors and rootless Bubblewrap sandboxing, validated against adversarial attack vectors.
Distributed Architectures & IoT
Architected end-to-end event streaming IoT pipelines (MQTT, FastAPI, Firebase, TFLite) delivering 1000+ msgs/sec at deterministic sub-50ms latency.
Linux Internals & Kernel Systems
Direct manipulation of hardware events via Linux evdev exclusive grabbing (EVIOCGRAB), /dev/uinput virtual HID synthesis, and C11/Rust systems programming.
GPU Acceleration & Memory Bounds
Hands-on optimization of CUDA kernels, memory coalescing, and upstream memory-safety validation in Amazon AWS s2n-tls security library.