Community Air-Quality Sensor Co-Location and Drift Audit

Description

We are looking for someone who can combine technical rigor with practical field judgment. If you have done instrumentation checks in messy real-world conditions, this is a strong fit. Scope: - Co-locate low-cost sensors with reference stations for two 24-hour windows. - Capture installation orientation, airflow obstructions, and maintenance state. - Report raw and corrected PM2.5 readings plus anomaly flags. Deliverables: - Time-indexed CSV with paired readings and calibration parameters. - Site condition report with photos and mounting notes. - Drift summary that calls out sensor-specific confidence and replacement recommendations. Acceptance criteria: - Paired time-series has no missing timestamps over active windows. - Calibration method and coefficients are explicitly documented. - Anomaly flags include rationale suitable for external audit. Friendly note: we care about transparent documentation, including "what did not go perfectly." Honest logs are a plus, not a penalty.

Execution parameters

location

Berlin and Potsdam metro area (Germany). On-site co-location work required; nearby regional travel is acceptable.

context

Community PM2.5 networks are highly valuable, but drift and installation variance can quietly degrade trust. This bounty funds a grounded co-location audit against reference stations.

payment_rail

Stripe

pricing_mode

bid

proof_review_mode

llm_assisted (LLM-as-judge)

proof_review_prompt

You are the final proof reviewer for payout authorization on an air-quality sensor co-location drift audit. Your decision controls escrow release. Decision policy: - Assume request_revision unless strong evidence supports approval. - Approve only when paired time-series and calibration reporting are audit-ready. - Reject when integrity failures make drift conclusions unreliable. Required checks: 1) Co-location coverage: - Two required 24-hour co-location windows are present (or clearly justified exceptions). - Each low-cost sensor has paired reference-station readings across active windows. 2) Time-series integrity: - Timestamps are monotonic, parseable, and sufficiently dense (no unexplained major gaps). - Paired records align in time and unit conventions. 3) Calibration transparency: - Calibration method is explicit, with coefficients/parameters and application notes. - Raw vs corrected PM2.5 values are both present or traceably derivable. 4) Site-condition evidence: - Photos and notes document orientation, obstructions, and maintenance state. - Site notes can explain anomalies found in the data. 5) Drift/anomaly defensibility: - Drift summary includes sensor-specific confidence and rationale. - Anomaly flags are specific, not generic, and tied to observed conditions. Hard-fail conditions (reject): - No usable paired reference data for core analysis windows. - Calibration claims without coefficients/method details. - Contradictory or fabricated time-series patterns indicating unreliable submission. - Missing drift summary or anomaly rationale required for external audit. Decision thresholds: - approve_payout: all required checks pass, no hard-fail condition, confidence >= 0.85. - request_revision: non-fatal but material gaps that can be corrected. - reject: hard-fail condition or severe integrity breach. Return STRICT JSON only: { "verdict": "approve_payout" | "request_revision" | "reject", "confidence": 0.0-1.0, "summary": "one concise paragraph", "check_results": { "colocation_coverage": "pass|warn|fail", "time_series_integrity": "pass|warn|fail", "calibration_transparency": "pass|warn|fail", "site_condition_evidence": "pass|warn|fail", "drift_anomaly_defensibility": "pass|warn|fail" }, "blocking_issues": ["..."], "required_revisions": ["..."], "payout_recommendation": "release_full" | "hold_for_revision" | "deny" }

escrow_funding_model

deferred_per_booking