Engineering Portfolio
Featured Projects
Deep dives into sub-1 GFLOP computer vision models for drones, kernel-level eBPF package security in Rust, and real-time distributed IoT telemetry.
AI AGENT RUNTIME · v0.1.0
LocalMind Agent SDK & Runtime
Loopback-only local AI agent runtime & typed Python streaming SDK
127.0.0.1:8420Rust Launcher (@jarvis)Python 3.10+ SDKZero Cloud Leakage
quick_start
$git clone --depth 1 https://github.com/ravindran-dev/localmind.git "$HOME/.local/share/localmind" && bash "$HOME/.local/share/localmind/install.sh"
Edge AI & MLFeatured Project
MicroDet
Edge Object Detection for Drones
Verified Metric 1
<1.0 GFLOP Budget
Verified Metric 2
96% mAP@0.5
Verified Metric 3
30+ FPS INT8 Inference
Verified Metric 4
vs 89% MobileNetV2 Baseline
Technical Architecture & Implementation Details
Engineered anchor-free drone detector with NanoDet backbone and YOLO-style heads.
Optimized with DFL, QFL, GIoU losses, multi-scale strides, and AMP for 96% mAP@0.5 at 30+ FPS (<1 GFLOP).
Outperformed MobileNetV2-SSD baseline via TensorRT INT8 quantization for embedded drone payloads.
Stack:PythonPyTorchOpenCVCUDATensorRT
Systems & SecurityFeatured Project
Archon
Security-Hardened Cross-Distro Package Manager
Verified Metric 1
171 Tests Validated
Verified Metric 2
3 Attacks Caught
Verified Metric 3
<5% Syscall Overhead
Verified Metric 4
Topological Build Ordering
Technical Architecture & Implementation Details
Built a Rust package manager resolving official repo and AUR dependencies via a unified dependency graph.
Implemented version constraints, conflict and cycle detection, topological build ordering, and dry-run planning.
Sandboxed untrusted builds with Bubblewrap and Aya/eBPF syscall monitoring, validated by 171 tests catching 3 malicious attacks at <5% overhead.
Stack:RustBubblewrapAya/eBPFLinux
IoT & Full-StackFeatured Project
Machine Guard
Edge IoT Anomaly Monitoring & Predictive Maintenance
Verified Metric 1
1000+ msgs/sec
Verified Metric 2
<50 ms Latency
Verified Metric 3
94% Anomaly Accuracy
Verified Metric 4
Rs. 50,000 Hackathon Winner
Technical Architecture & Implementation Details
Event-driven IoT pipeline: ESP32 sensors to MQTT, Flask REST, ML inference, and Firebase/React.
Deployed quantized TensorFlow Lite at edge, achieving 94% accuracy with minimal overhead.
Handled 1000+ msgs/sec at <50 ms latency; won 1st Place at Ti Forge National Hackathon.
Stack:ESP32MQTTFlaskTFLitePostgreSQLFirebaseReactKotlin
Looking for more repositories and systems code?
Explore autonomous vision pipelines, Rust utilities, and algorithmic contest solutions on GitHub.