All open roles

Knowledge

Regulatory Knowledge Engineer

Apply for this role

A conversation, a three-hour take-home on real regulatory text, then a session where we argue with your answers. Applications reviewed weekly.

About the role

Somebody has to read the Customs Act, the Foreign Trade Policy, the IRDAI broker regulations and a stack of notifications, and turn what they say into something a computer can evaluate without lying about it. That is this job, and it is harder than it sounds.

The difficulty is not comprehension. It is discipline — citing the span rather than the document, refusing to smooth over a provision that genuinely cannot be resolved on the facts available, and keeping all of it correct through an amendment cycle. Most people who read law well cannot do this. We are looking for the ones who can.

You do not need to write code and we will not test you on it. You do need to be comfortable with structure: decision tables, if-then conditions, fields with types, and the idea that a rule without an evidence requirement is not finished.

Backgrounds that work here are wider than the title suggests. Registered patent agents and law graduates from engineering or science backgrounds already draft and interpret at claim-level precision. Analysts from statutory compliance library firms have been decomposing acts into obligations with triggers, periodicity and evidence for years — under a different name. Tax technology and GST content people know exactly what breaks when a notification lands mid-year. XBRL taxonomy work, legal editorial with a house methodology, and trade compliance inside a manufacturer all transfer.

This is not a policy or advocacy role — you will not be writing about how AI ought to be regulated. It is not litigation, contract drafting or advisory work, and it will not keep your practice warm. And the prestige of your law school is not something we will weigh, because it does not predict the one aptitude we are testing for.

What you’ll do

  • Reading primary instruments — never a summary, never a commentary — and emitting typed rule objects with span-level citation.
  • Temporal correctness. Effective dates, supersession chains, and what the rule was on the date of a shipment rather than what it is today.
  • Evidence requirements. For each obligation: what document, issued by whom, bearing which field, would satisfy an auditor.
  • Saying "unclear" out loud. When a provision turns on a fact the system will never hold, that has to surface as a state rather than a footnote.
  • Keeping the rule base alive across amendments — the part that is invisible when it works and catastrophic when it does not.

What we look for

  • You work from primary text. If your instinct on an unfamiliar provision is to find a good summary of it, this is the wrong role.
  • You are comfortable with structure — decision tables, if-then conditions, schemas, typed fields.
  • You can recognise a provision that is genuinely undecidable on the available facts, and you say so instead of picking.
  • You will hold a position under pressure from a founder who disagrees with you. This matters far more than it sounds like it should.
  • You are precise in writing. Most of this job is written down.

Nice to have

  • Registered patent agent, or a law degree taken after an engineering or science degree.
  • Time inside a statutory compliance library, a tax technology content team, or a legal publisher's editorial operation.
  • XBRL taxonomy work, or any regulatory reporting standard with versioning and effective dates.
  • Customs, DGFT, foreign trade policy, SCOMET or marine insurance regulation specifically.
  • Any exposure to rules-as-code, decision modelling, or formal knowledge representation.

Application

Apply — Regulatory Knowledge Engineer

Twenty minutes. The written answers carry more weight than the CV, so give them the time.

A link is easier for us than an attachment — Drive, Dropbox, or your own site.

PDF only, up to 3 MB.

A few questions

The last part is the one we read closest. "Immediately" is rarely true.

0 / 1200

We are building a system whose job is to say "unclear" rather than guess. We are asking whether you do the same.

0 / 1200

The second half is the real question. Tax, safety standards, a sport, an insurance policy, a company travel policy — anything you know properly.

0 / 1800

We use what you send here only to assess this application. See our privacy policy.