Nikhil Ludder

<!-- Full-width GIF header -->

Nikhil — Systems Programmer & Full-Stack Architect

> *"I count bits to save a clock. I don't use valgrind; I know where I left my pointers. But when it's time to ship, I make the Cloud bend to my will."*

They say software is eating the world, but most of it is choking on bloated frameworks and electron apps. I prefer staying close to the metal. I write code that respects the CPU, compiles fast, and runs faster.

I'm a systems programmer, open-source maintainer, and now an enterprise-grade full-stack architect. Whether I'm writing bare-metal OS kernels, optimizing CUDA matrix multiplications, or orchestrating distributed containerized microservices on Google Cloud—I build systems that scale.

---

🔧 The Workbench

Here is what's compiling on my machine:

• The Todo — Distributed Task Orchestration. A massively over-engineered, full-stack microservice platform (FastAPI, React, Postgres, Redis). Features WebSockets, Celery background workers, and a full OpenTelemetry observability pipeline (Loki, Tempo, Prometheus, Grafana). Containerized and automated via GitHub Actions on a Google Cloud ARM64 VM.
• WebAI — Autonomous AI Orchestration. A native VS Code AI agent that bypasses API rate limits by reverse-engineering frontend APIs and using headless browser session-state emulation. Features an AST-aware unified diff patching engine and a fault-tolerant, 15-iteration self-correcting runtime.
• bindbc-cuda — Official Author & Maintainer. When I saw the D ecosystem relying on a dead CUDA library, I wrote a modern replacement from scratch for dcompute.
• MyOS — Wrote a 64-bit operating system from the UEFI bootloader up. Custom memory allocator, ACPI, xHCI (USB 3.0). Listed in Awesome OS because apparently, kids these days find writing kernels "awesome".
• CAMM (CUDA Matrix Multiplication) — A from-scratch FP32 SGEMM that outperforms NVIDIA cuBLAS for N ≤ 4096. Hand-tuned register tiling (asymmetric 16×8), float4 vectorization, double buffering, and an auto-dispatcher that picks the best kernel per shape.
• API Dash (GSoC 2026) — Gutting the core into pure Dart and building a multi-protocol (gRPC, MQTT, WebSocket) developer CLI. With Lots of PRs spanning core architecture refactors and model migrations.

---

⚙️ The Toolbelt

I don't chase trends. I use what works.

• The Old Guard: C, C++, x86-64 Assembly. (Where the real work happens).
• The Compute: CUDA. (If you aren't thinking in warps and shared memory tiles, you're just heating up the room).
• The Full-Stack & Cloud: Python (FastAPI), TypeScript / React, Next.js, Docker, Nginx, Google Cloud (GCP). (Deploying containerized microservices and automated CI/CD pipelines at scale).
• The Day Job: Flutter / Dart. (Because shipping Extremely Optimized e-commerce apps at Johar Basket and building production infra as CTO at Doordripp pays the bills).

---

📡 The Specs

• ☁️ Full-Stack DevOps Architect — Mastering Docker, Nginx, CI/CD, and OpenTelemetry to build resilient, auto-scaling SaaS platforms.
• 🔬 Google TPU Research Cloud — Accessing heavy iron for high-performance compute.
• ☁️ AWS Certified Solutions Architect — I know how the "cloud" works (it's just someone else's Bash).
• 🎯 GSoC 2026 Contributor — API Dash. Implementing low-latency multi-protocol support (gRPC, MQTT, WebSocket), and engineering a unified cross-platform CLI.

---

📊 Telemetry

---

"Any fool can write code that a computer can understand. Good programmers write code that humans can understand... and great programmers write code that doesn't waste my L1 cache or crash my Docker containers."

Collapse