What can be inferred without converting uncertainty into false precision?
Pearl Jin
What can be known from incomplete evidence?
I build computational systems that have to remain correct when evidence is incomplete.
Currently building a new company.
01 / Questions
Which assumptions determine the answer before computation begins?
When should a system refuse to answer?
What happens when error compounds rather than averages out?
How should confidence change when the model itself may be wrong?
Which failures come from noise, and which come from the structure of the problem?
What is lost when an estimate is mistaken for an observation?
How can correctness survive incomplete evidence?
02 / Principles
Working under incomplete evidence.
01
Evidence before confidence
A precise answer is not necessarily a supported answer.
02
Assumptions are part of the system
Hidden premises often determine failure long before a model produces an output.
03
Refusal can be correct
A system that cannot distinguish knowledge from guesswork should sometimes remain silent.
04
Error has structure
Failure is not always random. It can accumulate and reinforce itself.
03 / Research Notes
Short technical memoranda on computation, systems, and numerical methods.
04 / Selected Systems Work
Technical work from different problem domains.
Deterministic Computational Systems
Replayable execution, constraint validation, immutable provenance, idempotent effects, and numerical reproducibility.
Aliqubit is an open-source systems project concerning deterministic and constraint-aware computation.
Browser Privacy Infrastructure
Local-first privacy boundaries, permission minimization, DOM observation, client-side detection, and redaction before data reaches an external service.
Structured Knowledge Systems
Semantic retrieval, explicit relationships, provenance, temporal context, and representation of conflicting claims in personal knowledge systems.
05 / Resume
Pearl Jin
Founder and engineer. Attended the University of Southern California; LLB studies at Monash University. Software engineering learned through practice.
Work across security, simulation, optimization, and systems. Recurring concern: correctness when information is incomplete and assumptions fail.
Currently building a new company.