Senior Backend Developer (Python, PostgreSQL) - Part-time - LATAM/PST

Role Overview

Company: NūR Scientific
Focus: Backend-first, minimal UI surfaces, epistemic integrity under pressure.
Core requirement: Zero architectural reinterpretation. This is a transcription build from locked artifacts.
Timeline: 3–6 weeks (fixed scope, acceptance-test driven).

We are hiring a senior backend engineer to implement a tightly scoped “vertical slice” of an epistemically constrained system.

The architecture and execution surface are finalized. Your role is to implement exactly what is specified and pass predefined acceptance tests. This is a transcription build, not a product design role.

Creativity, architectural reinterpretation, feature suggestions, or speculative improvements are explicitly out of scope.

What You’re Building

• A subject-scoped, multi-actor system that enforces strict authority and audit guarantees:
• Human user: sole authority to accept/reject meaning
• AI actor: proposer only; cannot create or promote canonical records
• Explicit acceptance model: binary, immutable, inspectable
• Full attribution + provenance for all meaning claims
• Append-only audit history
• Lexical recall only (no ranking, synthesis, inference, or correctness scoring)

The vertical slice must prove in running code that:

• No meaning becomes canonical without explicit human acceptance
• No silent state transitions occur
• Every proposal, acceptance, rejection, and forbidden action is auditable

System Constraints (Non-Negotiable)

Anything not explicitly permitted is forbidden.

The build must not include:

• Multi-user acceptance or collaboration workflows
• Consensus systems
• Semantic graphs, ontologies, or domain packs
• Confidence/correctness scores
• Ranking or recommendations
• Background automation or auto-promotion
• Embedding or semantic search (lexical only)
• UX polish beyond minimal functional surfaces
• “Future-proofing” fields or inactive scaffolding that enables forbidden behavior

Where the spec or tests are silent, default behavior must be:
Explicit rejection or no-op
Not inference or reasonable defaults.

Acceptance tests are authoritative.

Core Objects (Persisted)

• Document (subject-scoped, raw artifact + provenance)
• Entry/Event (human-accepted canonical record)
• Interpretation (AI-proposed meaning)
• DerivedSnapshot (non-canonical derived output)
• AuditLog (append-only, immutable)

Required Behaviors

• Every write is scoped to exactly one subject
• If subject not provided → DEFAULT_SUBJECT (“general”)
• AI may propose interpretations and derived snapshots
• Human may accept/reject interpretations exactly once
• Acceptance state transitions are immutable
• Forbidden writes are synchronously rejected (no queuing, retrying, or downgrading)
• Audit history must reconstruct “who believed what, when, and why”
• Extraction is allowed for convenience, but no normalization, ontology mapping, classification, or canonical structuring occurs without explicit human action.

Tech Stack

Flexible, provided acceptance tests pass.

Typical fit:

• Backend: Python (FastAPI) or equivalent
• DB: PostgreSQL (preferred) or equivalent relational DB
• UI: Minimal surfaces only (upload, proposal review, accept/reject, audit view, lexical search)

Engineering Expectations

• Enforce hard gates around actor permissions and immutability
• Strong relational schema discipline (no speculative fields)
• No placeholder logic that enables forbidden behavior
• Provide a focused test suite proving invariants
• Deliver clean repo, deterministic local setup, and run instructions

Ideal Profile

Senior backend engineer

• Experience in regulated, ledgered, or audit-heavy systems
• Comfortable implementing strict invariants
• Will halt on ambiguity rather than infer
• Values constraint over creativity

Definition of Success

A hostile reviewer can verify that:

• The system cannot silently decide or drift
• AI cannot promote or canonize meaning
• Every state transition is attributable and immutable
• All meaning claims are acceptance-gated and auditable