We are process open, not just outputs.

Not a bug. A feature of how food data works — or doesn't. The system had ingredient lists but no way to know what they actually meant. Maida and refined wheat flour are the same thing. Kashmiri chilli is not just a noisy version of chilli. String matching cannot tell the difference. Until a system can, nutrition studies run on shaky ground, compliance checks guess, and a regulator cannot ask a clean question and trust the answer.

The coordination layer that makes these questions answerable does not exist for Indian packaged food. Building it is what we are doing.

That work has produced things we did not expect going in.

A classification framework grounded in ITC-HS codes, FSSAI regulations, and Supreme Court rulings — inspired by S.R. Ranganathan's 1933 Colon Classification — that assigns ingredient identity across three axes: energy profile, material composition, and label function. The first deterministic fully backtracable framework that quantifies NOVA processing at the ingredient level, not the product level.

A sampling methodology for what we call fragmented textual spaces — domains where no ground truth exists and the vocabulary itself is part of what you are trying to map. Google Research independently converged on the same structural problem from flood data. The methods note generalises the approach so other domains can use it.

A comparative analysis method — Regulatory Delta Analysis — that reads two versions of a law not as text but as data: what changed, what stayed, what the delta reveals about the constraints the system is working within. Applied so far to FSSAI 2011 vs 2020 and to a century of landmark rulings from Supreme Court that redefined food classification hierarchy.

896 SKUs. 2,291 ¹ unique ingredient variant strings. All of it open under CC BY 4.0, process and dead ends included, on GitHub.

Sponsorship keeps the infrastructure running. We are open, not just in the final outputs being open access - but the process, dead ends, debates and learning, all of it happens in the open in GitHub.

Ya Devi Sarva Bhuteshu Buddhi Roopena Samsthita, Namasthasyai Namo Namah - Devi Mahatmiyam¹

Salutations to the Goddess who dwells in all beings in the form of intelligence.

The i in iSRL is lowercase on purpose. The work is not about any single person. Everyone here serves the same thing.

iSRL is an open research lab. We pick one focus area per year, go deep, and publish everything — findings, dead ends, process, questions we couldn't answer. All of it, openly, under CC BY 4.0.

The work sits at the intersection of systems, infrastructure, and society — problems where the gap between how things are organised and how they could be organised has real consequences for real people. We build the substrate that lets other people ask better questions.

A coordination layer for India's food systems. Thousands of regional ingredient expressions — Mathania red chilli, Kashmiri mirch, lal mirchi powder — mapped to stable, machine-readable identifiers without any one of them being declared the correct name. Brands keep their identity. Regulators get a consistent reference. Researchers get reproducible data.

Coordination without convergence. The standard lives in the infrastructure, not at the surface.

Active sprint → Allergen Declaration in India — Rules, Reality, and Gaps

The entry point is the work. Browse the open research threads — if something catches your eye and you have something real to say, drop in. That is how most people find their way in.

If you are a subject matter expert, a researcher, or someone with domain depth in food systems, regulatory law, nutrition science, data infrastructure, or adjacent fields — the lab engages with advisors and collaborators on specific problems, not as a formality but because the work genuinely needs people who know things we don't. If something in the research threads connects to your domain, that connection is the invitation.

If you are new to research entirely — also welcome. The threads are designed to be legible to someone who has never written a research report. The milestone structure means you are never asked to produce everything at once. You start where you are.

Active threads → sandbox-research discussions

All outputs on Zenodo → All Research Logs →

Fully remote, fully async. No office, no fixed hours. What we keep is a deliberate rhythm — every third week completely off, high agency on work weeks, and genuine support through