Ajith Sahasranamam Columbus, Ohio

Ajith Sahasranamam

I combine ideas from different fields and use AI to put them to use.

For that to work, AI has to be dependable. So I work on it from both sides: building with AI, and making AI that can be depended on.

Now

Reading George Musser’s Putting Ourselves Back in the Equation, a library copy, for the next essay.

Building with AI

2 November 2026
0:00 / 5:17

Counted When Promised

How much of rich countries’ climate finance actually arrives?

Five minutes on the headline climate-finance total, which mixes promises with payments, and what it would take to check each promise against what was paid. Every figure on screen names its source.

Made on one machine with AI tools. The narration is my voice, cloned.

The Substack post has a short note on what the film covers.

Since October 2026

Climate Finance Commons

A public record of how each country reads its climate-finance obligations. Each reading is written as a formal model, so anyone can check what it commits that country to. It starts with one investigation.

In progress

Making AI dependable

Since 2016

Ongil

The company I founded. It tests what an AI agent is allowed to do before the agent acts inside a bank or an insurer.

4 March 2026

Accountability Theater

An essay. The deliverable for the software engineer of 2026 is accountability, and the leader’s job is to make that possible.

Next

An essay that starts from Musser’s book and ends at testing AI agents. On Substack when it is ready.

Filed with NIST

Sent on behalf of Ongil

23 March 2026

Comment on the NIST AI 800-2 draft

On practices for automated benchmark evaluations of language models. Acknowledged by CAISI staff, who said it will be considered in the revision.

14 March 2026

Barriers to AI adoption in finance: build-time compliance

Input to CAISI’s finance listening session. Acknowledged by CAISI staff.

Research

Computational neuroscience

March 2020

Uncoupling the roles of firing rates and spike bursts in shaping the STN-GPe beta band oscillations

In Parkinson’s disease, a circuit deep in the brain locks into a strong 15 to 30 Hz rhythm, and its neurons both fire faster and fire in bursts. Since the two changes arrive together, nobody could say which one drives the rhythm. We separated them in a simulation of 3,000 neurons. How fast one group of neurons fires decides whether the rhythm appears. Bursting matters only when the circuit sits near the edge of rhythm, where it can tip it either way. That may explain the short bursts of rhythm seen in recordings.

With Jyotika Bahuguna and Arvind Kumar. PLOS Computational Biology.

May 2016

Dynamical state of the network determines the efficacy of single neuron properties in shaping the network activity

Neurons differ in how they fire: some at a steady pace, some in bursts. We built a model neuron whose firing pattern can be changed without changing how much it fires, and asked whether those patterns change what a whole network of neurons does. The answer depends on the state the network is in. Across most states the patterns make no qualitative difference. Near certain tipping points they can start or stop a rhythm. When each neuron’s pattern also depends on the network’s activity, the network’s response to an input comes to depend on its recent history.

With Ioannis Vlachos, Ad Aertsen and Arvind Kumar. Scientific Reports.