Hi, I’m Srinidhi Ramanujam, a consulting technologist based in Bangalore.
I bridge the gap between enterprise finance and applied AI, building agents that work and platforms that last.
This site is a log of my technical deep dives, experiments curiosities, and the odd bit of fiction.
Currently focused on#
- Building and measuring LLM workflows that ship
- Exploring retrieval and agent patterns with real constraints (cost, latency, failure modes)
- Logging the experiments, lessons, and reversals - and the odd bit of fiction
Explore the site#
- Writing: Long-form essays on data platforms, AI,
engineering leadership, and the systems thinking behind them.
- Fiction: Narrative experiments that keep my creative
muscles honest.
- About: A concise bio and how I approach this work.
Spotlight#
Spotlight
Jan 2, 2026 · Writing
I’m going to say something that might sound reckless: In 2026, “Human-in-the-Loop” may become the new “Red Flag Act.”
In 1865, the UK required a man to walk 60 yards ahead of every automobile, waving a red flag to warn pedestrians. It wasn’t about safety, it was about comfort. It made the new technology legible to our intuition. And it may have held back the British car industry back for 30 years.
I’m going to say something that might sound reckless: In 2026, “Human-in-the-Loop” may become the new “Red Flag Act.”
In 1865, the UK required a man to walk 60 yards ahead of every automobile, waving a red flag to warn pedestrians. It wasn’t about safety, it was about comfort. It made the new technology legible to our intuition. And it may have held back the British car industry back for 30 years.
...
After 4,380 API calls, we found that context engineering matters more than context size.
Testing whether million-token context windows actually beat disciplined retrieval and packaging.
2025 was supposed to be the year of AI agents. Most faded into slideware. Coding agents are the exception. Here’s how I actually use them.
The Age of Directives
by Srinidhi Ramanujam
The Academy of Applied Directives was less a university than a cathedral. It was the highest temple of knowledge in the Federated Metropolis.
It did not train students to work, because work was unnecessary.
The Engines had made sure of that. These were not machines in any familiar sense. They were planetary-scale minds: distributed networks of computation and knowledge models built up over centuries, refined generation after generation until they could do anything asked of them. They fueled everything.
...
Originally written in June 2025. This is a more detailed version of a talk from around the same time.
Everyone keeps saying “AI is changing software engineering.”
Depending on who you listen to, that means either:
“Developers are finally 10x more productive,” or “Developers are finally 0x employed.” Reality, as usual, is less dramatic and more interesting.
A normal day for most engineers in 2025 looks something like this:
You open your IDE. You have a TypeScript backend, or a Java microservice, or a Python data pipeline. There is an issue tracker, a CI pipeline, some dashboards that may or may not be lying to you. Someone on the team is arguing about naming and someone else is quietly fighting with YAML.
...