Most of this is invisible from the outside, so here it is plainly: not a keyword list, but the surface I've actually built on and the judgment that came from shipping on it.
I work at the frontier of these tools, and bring teams along with me.
On this since the first wave: ChatGPT and GitHub Copilot through Anthropic's Claude Code and its Max tier. Usually first to new tooling, and the one who helps the team adopt it.
Parallel AI agents are my normal way of working. I instrument my own usage, thousands of sessions of it, to keep getting sharper.
RAG pipelines, MCP servers that hand agents real tools, agentic workflows, local and open models (Ollama), and the wider ecosystem.
I track and test what's new and turn it into better ways of building. It's the engine behind every product I ship.
Web and mobile front-ends through backend microservices. I own the whole stack, so nothing falls through the seams between teams.
Analysis and processing pipelines built to run at the scale of terabytes. That's the volume where the difference between a script and an architecture shows up in the bill.
With Temporal I break a big process into independent pieces, each scaled, retried, and reasoned about on its own. Durable, and far easier to scale and maintain.
High-end GPU compute (H100 / A100) tuned for efficient utilisation; network and database costs optimised on heavy-data infrastructure. Fast is half the job; cheap at scale is the other half.
Fleets of agents and concurrent pipelines instead of one thing at a time. That's how a serial document path became 30–40× faster on the same hardware.
Retries, idempotency, and recovery paths so a hiccup doesn't become an outage. The system keeps its footing when an upstream call flakes or a document fights back.
Containerised with Docker, orchestrated on Kubernetes, and hands-on across AWS, GCP, and Azure. The system is never hostage to one vendor's pricing.
Kubernetes runners that autoscale from zero, plus an always-on Mac Studio (M2 Ultra, 192GB). My own CI fleet; builds and tests run on my terms.
Production monitoring with Grafana dashboards and live metrics. Systems designed for both horizontal and vertical scaling, and the judgment to know which one a problem needs.
Code-signed, notarised macOS apps in Swift and SwiftUI, with OAuth device flow and auto-update, shipped to real machines. Most teams need two people for this. I'm both.
Agentic workflows that carry a task from start to finish, backed by automated end-to-end testing, so the automation can be trusted with the boring, critical work.
End-to-end automation that takes thousands of manual hours out of a business. Unattended pipelines that remove humans from the loop for good.
The same instinct turned on my own work: research, tooling, dependency and security checks across thousands of repositories, my own sessions instrumented. One person operating like a team.
I don't accept how something is currently done. I research a better way, benchmark it, and rebuild. That's how the legacy systems got beaten: not tuned, replaced.
This is the engine behind every product. See what it built.
This site is my own work, end to end — like everything it describes. · v1.3.0 · cf9ec25
Hey, good to meet you. Quick heads up: I'm an AI version of Rahul, trained on his work, so I can talk through what I build and how I work any time. What brings you by? Got something you're trying to get built?