Hellomy name is

Ishan Ambasta

I solve problems at Alt Carbon.

Alt Carbon · Bengaluru, India

Portrait of Ishan Ambasta
Bengaluru
From Ishan AmbastaRe about me

I’m Ishan Ambasta. I work on AI and internal tools at Alt Carbon, a company that removes carbon dioxide from the air by spreading crushed basalt on farmland.

Most of my work is making software that people who don’t care about software can use without help. At the moment that means The Portfolio Project: every colleague gets a personal site they can edit themselves and take with them when they leave.

I like websites that are small, fast and made of text. This one is my argument for them.

Work

  1. 2023 – now Software lead, Alt Carbon
  2. 2020 – 2023 Software engineer, FieldKit Labs
  3. 2018 – 2020 Data analyst, Monsoon Analytics

Now

Work

  • Building The Portfolio Project: a personal website for everyone at Alt Carbon, which they keep when they leave.
  • Prototyping three plain-text templates. You are looking at one.
  • Testing Sveltia CMS as the editor and Cloudflare Workers as the host.

Reading

  • Basin and Range, slowly.
  • Rereading The Checklist Manifesto, because of a deployment checklist.

Life

  • Long walks, fewer tabs.

Not doing

  • New side projects, until this one ships.

updated 14 Sep 2026

September2026

MoTuWeThFrSaSu
123456
78910111213
14151617181920
21222324252627
282930

↑ last updated the Now note

On the nightstand

Fetching the shelf…

Reading list

What I’ve been reading, and what I’d pass on.

No. 011 · all items keep

  1. How to Do Great WorkPaul GrahamThe one I send to new hires.
  2. An app can be a home-cooked mealRobin SloanSoftware for four people is allowed to exist.
  3. A Brief History & Ethos of the Digital GardenMaggie Appleton
  4. FastPatrick CollisonA list of things built quickly. Cheering.
  5. You and Your ResearchRichard HammingA 1986 talk that still reads like advice from a colleague.
  6. The Tail EndTim Urban
  7. Potential for large-scale CO₂ removal via enhanced rock weathering with croplandsBeerling et al., Nature, 2020The paper most of our field is built on.
  8. Inventing on PrincipleBret Victor, 2012Why tools should show you what they’re doing.
  9. VoltsDavid RobertsLong interviews about the energy transition.
  10. Works in ProgressMagazine
  11. CDR.fyiCarbon removal market dataWhere I check who bought what.

Items11

Thank you, come again

Research

  1. Preprint · EarthArXiv 2026An offline-first pipeline for soil sampling in enhanced rock weathering trialsMeera Iyer, Asha Rao, Daniel Okafor, Priya Nair
  2. Dataset · Zenodo 2025Soil cation measurements from basalt-amended rice paddies, 2024–2025Daniel Okafor, Asha Rao, Meera Iyer
  3. Talk · Field Data Workshop, Bengaluru 2025Designing forms for people who don’t want to fill them inAsha Rao
  4. Report · Alt Carbon technical report 2024Measurement, reporting and verification for enhanced weathering: a practitioner’s checklistPriya Nair, Meera Iyer, Asha Rao, Rahul Menon

Writing

  1. A website you keep when you leave2026Why every colleague should own their site, account and all.
  2. Plain HTML is a feature2026
  3. Forms are the product2026Most internal software is a form. Treat it like one.
  4. Offline-first forms for field teams2025
  5. What a carbon tonne is made of2025Rock, rain, soil samples and a lot of spreadsheets.
  6. Tools nobody asked for2025
  7. Boring softwareexample.com2024
Call Ishan ishan@altcarbon.com

Say hello

Email is best: ishan@altcarbon.com. I answer most things within a week. I’m always glad to hear from people building small tools for field teams.

Pin back

21 Aug 2025 · 450 words

What a carbon tonne is made of

A tonne of removed carbon is a claim backed by measurements. When someone buys one, they are buying a chain of evidence that starts in a field and ends in a spreadsheet. This is a tour of that chain, from the software side.

The field

Basalt dust goes onto farmland. Rain dissolves it slowly, and the reaction pulls carbon dioxide out of the air and carries it away as dissolved carbon. Nobody can watch that happen, so we measure around it:

  1. What went onto the field: how much rock, from which quarry, of what composition.
  2. What the soil looked like before: samples from fixed points, taken before spreading.
  3. What changed after: the same points, sampled again across seasons.

Each step produces paperwork. Most of the paperwork used to live in notebooks.

The sample

A soil sample is a bag of earth with a label. The label is where most of our problems started. A smudged date, a field name spelled two ways, a point number written down from memory that evening. The lab result is only as good as the label it comes back attached to.

A measurement you cannot trace back to a place and a time is not a measurement. It is a number.

So the first tool we built was not clever. It is a form that runs on a phone with no signal, fills in the field and point from GPS, and prints a label. The person in the field types almost nothing.

The spreadsheet

Results come back from the lab as files. We join them to the sample records and check them before anyone does any arithmetic:

Check Why it matters
Every result has a sample Orphan results are usually typos in the label
Every sample has a result Missing results are lost bags or lab backlog
Values sit in a sane range A misplaced decimal looks like a miracle field

The joining step is short enough to show:

results = lab.merge(samples, on="sample_id", how="outer", indicator=True)
orphans = results[results["_merge"] != "both"]

Only after the orphans list is empty does anything go to the scientists.

The tonne

What a buyer finally sees is one number with a range around it.1 Behind it sit a few hundred bags of soil, a few thousand form submissions, and a long list of small corrections nobody will ever read. That list is the product. The number is a summary of it.

Footnotes

  1. The range is not a disclaimer. It is the honest part of the claim, and it gets narrower as the evidence improves. ↩

Reading

What I’ve been reading, and what I’d pass on.