Global Trade Analyst - SME/QA

Job description

**Goods SME / Labeling & QA Analyst (6 Months Contract)** **Role Summary** We need a person who can read messy goods descriptions (e.g. bills of lading, invoices, air waybills, internal cargo manifests, MT 700), evaluate the Goods Engine’s extracted goods, then validate the engine’s screening outcome by labeling results as True Positive / False Positive/True Negative/False Negative (and why) with strong knowledge of controlled goods, export controls, dual-use items, and luxury goods risk in the context of trade flows. **Primary Tasks** - Interpret goods descriptions (input understanding): - Read and interpret real-world, imperfect goods text from the data (from Financial Institutions, Air Cargo Operators, Freight Forwarders etc.) - Normalize descriptions into what the goods are, including: - Synonyms, abbreviations, misspellings, HS-code-like phrasing, multi-item lines - Brand + model references (SFK (brand) ball bearing) - Ambiguous descriptors (“spares”, “parts”, “equipment”, “machinery”) **Why it matters** : This is the foundation for judging whether extraction + screening is reasonable. **Review Goods Extraction from the Goods Engine (extraction QA)** - Compare the goods and attributes extracted by the engine against the original description. - Review potential relevance under EAR, ITAR, USML, Commerce Control List, UK strategic export controls and EU dual-use and military controls. - Recognize when additional technical documentation is required - Identify issues like: - Missing key noun/ commodity (“controller” extracted but not “servo motor”) - Wrong granularity (extracting “parts” instead of “gyro-stabilized drone gimbal”) - Incorrect segmentation (combining two items into one or splitting incorrectly) - Incorrect mapping to controlled/dual-use categories due to ambiguous terms **Output** : Structured extraction QA notes: “extraction correct / partial / wrong” with reasons. **Review engine screening results + label outcomes (TP/FP with reasons)** - Evaluate the engine’s outcome and label: - True Positive (TP): engine correctly flagged controlled/ dual use/ luxury goods risk - False Positive (FP): engine flagged but it’s benign / misinterpreted - (Optional but recommended) False Negative (FN): engine missed a controlled/dual use/luxury indicator - Provide reason codes and evidence: - Keyword triggers vs. contextual disambiguation - Misread abbreviations (e.g. AR for “augmented reality” vs “assault rifle” ambiguity) - Commodity vs. end-use confusion (“laboratory centrifuge” vs. “uranium enrichment component”) **Output** : Consistent labeling decisions + rationale that can be audited and used for model/engine improvements. **Ideal Candidate Profile** **Domain Knowledge (must-have)** - Strong understanding of: - **Controlled goods (weapons, ammunition, military components, restricted chemicals)** - Export-control classification - EAR and Commerce Control List concepts - ECCNs and EAR99 - ITAR and US Munitions List concepts - **Dual-use goods** (electronics, sensors, navigation components, UAV/drone parts, telecom gear, machine tools, high-spec materials) - **Luxury goods** signals (high-end watches, jewelry, designer goods) and how they appear in trade docs - **Trade, shipping and logistics documentation** patterns and shipping descriptions (air/sea cargo phrasing) **Nice-to-have:** - Familiarity with **HS codes** , commodity classification, trade compliance screening logic - Exposure to **export controls** concepts (even if not a legal specialist) - Exposure to practical goods classification and export-control experience, strong knowledge of at least one major jurisdiction—preferably across the US, UK and EU—and the ability to interpret complex goods descriptions and validate screening engine outputs with technical teams. **Analytical & Labeling Discipline** - Can follow a labeling consistently (reason codes, evidence, confidence) - Comfortable with high-volume review work without drifting in standards - Excellent written rationale (audit-friendly, regulator-safe language) **Technical Comfort (not mandatory)** - Comfortable using: - Ticketing / workflow tools (Jira/Confluence) - Simple annotation tools or spreadsheets - Reviewing “engine output” fields (extracted entities, confidence scores, matched rules/contexts)