+--------------------------------------------------------------+
| 🔴 🟡 🟢 arpan@pathak — zsh |
+--------------------------------------------------------------+
| |
| ~ $ whoami |
| Arpan Pathak |
| Systems Software Engineer · 8+ yrs |
| Rust · Go · C++ · Kotlin · Python |
| Amazon → Oracle → Microsoft |
| |
| ~ $ motto |
| "Never settle, always regret, |
| then refactor the mess." |
| |
| ~ $ wisdom |
| "Age gele baaghe khai" |
| আগে গেলে বাঘে খায় |
| (rush ahead and the tiger eats you) |
| |
| ~ $ _ |
+--------------------------------------------------------------+
🛠 Languages
⚙️ Systems · Edge AI · Cloud
$ whoami
Systems engineer, 8+ years: Amazon → Oracle → Microsoft.
I make databases faster (lock-free), networks safer (eBPF/Cilium),
and ML inference cheaper (100ms pipelines, 5k QPS serving).
$ cat interests.txt
Rust • Systems Programming • Distributed Systems • Computer Vision
LLM Inference • Game Physics • Computational Geometry • Kubernetes
Local-First AI • Agentic Coding • Edge Devices • Open Source
$ philosophy
"Never settle, always regret, then refactor the mess."
"Age gele baaghe khai" — আগে গেলে বাঘে খায়🦇 Now Building: openBatarangs
Agentic coding CLI for local models. Auto-discovers the best Ollama model on your hardware, then reads, edits, and verifies code — no cloud, no manual model ops. Built for Jetson-class edge devices and every laptop that runs Ollama.
🚗 Now Building: Driving CivicSense
Edge-AI co-pilot for road safety. 100% on-device perception, privacy-first, zero video leaves the car. Runs on NVIDIA Jetson Orin with TensorRT.
| Module | What it does |
|---|---|
| driving-civicsense-vision-model | Main repo: YOLOv8/v11 INT8 → TensorRT (~12 ms), kinematic decision engine, Kalman + Deep SORT, voice alerts, research paper + ISO 26262 safety study |
| civicsense-pi-stream | Capture layer: Rust MJPEG + UDP live dash-cam on Raspberry Pi Zero 2 W — hotspot mode, ~50 MB RAM |
| civicsense-stream-client | Perception layer: pure-Rust YOLOv8n with HuggingFace Candle — no Python, no ONNX Runtime |
| civicsense-companion | Human layer: Kotlin Multiplatform app (Compose + SwiftUI) with gRPC real-time alerts |
| driving-civic-sense-data-crowd | Data layer: MIT-licensed 7-class YOLO labels, ground-truth validators, public-dataset aggregation |
🦀 New: cuda-oxide-demo
Idiomatic CUDA kernels written in Rust with CUDA-Oxide, benchmarked on Jetson Orin: SAXPY, dot product, tiled matmul, and a Laplace/Jacobi PDE solver. Includes a detailed article and a GitHub Pages book chapter.
| Link | Where |
|---|---|
| cuda-oxide-demo | Main repository |
| CUDA_RUST_JETSON_BENCHMARKS.md | Detailed article |
| GPU Parallel Programming Book | Book chapter: Benchmarking CUDA-Oxide on Jetson Orin |
- 🔭 Now building: openBatarangs — local-first agentic coding CLI for Jetson & Ollama · Driving CivicSense — edge-AI road safety · grungeGPT — educational from-scratch GPT in Rust
- 🌱 Deep-diving: CUDA-Oxide, eBPF/Cilium, edge LLM inference, lock-free systems
- 💬 Ask me about: Rust, systems programming, GPU kernels, local LLMs, Kubernetes
- 📫 Reach me: arpan.pathak47@gmail.com
| Layer | Weapons of choice |
|---|---|
| Languages | Rust, C++, Go, Kotlin, Python, Java, TypeScript, C |
| Systems | Lock-free data structures, memory safety, kernel/eBPF, low-latency |
| Networking | TCP/IP, UDP, TLS/mTLS, HTTP/2/3, gRPC, QUIC, Cilium, XDP |
| Distributed | Kafka, Flink, Spark, consensus/replication/sharding, CDC |
| ML/AI | PyTorch, TensorFlow, BERT, ONNX Runtime, Ollama, edge inference, CUDA |
| Cloud | AWS, GCP, Azure, Kubernetes (EKS/AKS), Terraform, Kubebuilder |
|
🦇 openBatarangs Rust · local agentic coding CLI ▶ watch demo |
⚡ gpu-parallel-book CUDA · GPU · Rust · 16-chapter book |
🚗 civicsense-vision-model Rust · on-device AI perception |
|
📚 seeing-machines-book Python → Rust · CV book |
🐘 multicloud-distiller Rust · distill LLMs on 6 clouds |
🎮 bevy-physics-book Rust · physics in Bevy |
|
☸️ kube-goAt Go · Kubernetes IaC |
🎛️ andross C++20/23 · low-latency Vision/AR SDK |
🔍 systems-patterns Go ⭐ · systems hands-on |
|
🔐 zero-trust-networking Go · Cilium · K8s zero-trust |
🥊 FightClub_600 Interview prep · 5 languages |
🧠 AdvancedAlgoPatterns Kotlin · DSA patterns |
|
🦀 jobsense-parker Rust · job-hunt CLI |
📘 modern-cpp-systems-book C++ · memory-safe systems book |
🕸️ ebpf-rust-networking-book Rust · eBPF/Cilium/K8s book |
|
🎸 grungeGPT Rust · from-scratch GPT + BPE + GGUF/Ollama |
🦀 cuda-oxide-demo Rust · CUDA-Oxide · Jetson CPU-vs-GPU benchmarks + Laplace/Jacobi solver book chapter |
|
…plus 40+ more repos: distributed systems, ML notebooks, music & audio, websocket scaling, QUIC tools, NFC + e-ink hardware, and a Neovim agentic IDE.
🏆 Flagships: openBatarangs (local-first agentic coding CLI for Jetson + Ollama) · driving-civicsense-vision-model (edge-AI road safety, 100% on-device, research paper + submodule pipeline) · gpu-parallel-book (CUDA, GPU & parallel programming, Rust + CUDA-Oxide, 16 chapters) · seeing-machines-book (computer vision, Python → bare metal) · deepseekcoder-multicloud-distiller (distill LLMs across 6 clouds with full data sovereignty)
|
🚿 bathroom-singing-cli TUI studio · bathroom vocals → production |
🌀 nirvoid Rust · audio processing library |
🎧 audio-stem-splitter Rust · vocals → stems (Demucs) |
|
🎼 gen-music-models generative ML · music, art, poetry |
🎛️ audio-ml audio EDA · ML models |
➕ more repos systems · ML · hardware |
| Company | Role | Impact |
|---|---|---|
| Microsoft | Software Developer II | Led legacy C++/C# → Rust migration: 40% lower P99, 2× throughput, 3× less memory, 6-figure savings. Built AI policy engine + zero-trust sovereign networking (eBPF/Cilium). |
| Oracle Cloud | Senior Software Engineer | Lock-free DB internals: +45,000 ops/s. Terraform provider for Autonomous DB (~$1.2M/mo savings). |
| Amazon | SDE II | Real-time ML ranking (XGBoost + DL): 100ms latency, 5% CTR lift, $5.4M/mo revenue. BERT inference @ 5,000 QPS. Petabyte-scale data platform. |
| Razorpay | SDE | Go UPI gateway: 1M+ hourly transactions, 99.99% uptime. |
LinkedIn · YouTube · GitHub · Portfolio · 📧 arpan.pathak47@gmail.com
Currently: engineering systems that make AI fast, safe, and affordable — building openBatarangs for local-first agentic coding — and writing books about how it all works.