AI Engineer (Adaptive Feedback & Assessment/ Learner Graph & Recommendations)

AI Engineer #1 - Adaptive Feedback & Assessment
Mission: Own the brains behind Genie's instant feedback: deliver marking, commenting, and note-summary models that feel human, hit ≥ 95 % accuracy, and answer in
Key Responsibilities:
● Train & fine-tune NLP / vision models for Math & Science auto-marking, AI commenting and class-note summarisation.
● Build retrieval-augmented explanation pipelines so every hint cites workbook lines or teacher notes.
● Ship fast: instrument latency, automate regression tests on 2 k+ sample sets, and roll out safely via feature flags.
● Work with Curriculum to label edge-case data and measure real-world accuracy weekly.
● Add guard-rails - bias checks, "double-check with a teacher" fallback when confidence dips.
Success Looks Like:
● 95 %+ exact-match score on marking benchmark; p95 latency
●
● Models & infra ready for staged roll-outs (staff → beta → all) with zero downtime.
AI Engineer #2 - Learner Graph & Recommendations
Mission: Design the intelligence that makes Genie a "smart-cool friend" - a mastery graph, progress road-maps, and streak-goal engine that boost weekly practice by +20 %.
Key Responsibilities:
● Model concept mastery and temporal learning gaps in a graph database (Neo4j / Amazon Neptune).
● Build recommender services that output next-best-question sets and personalised streak targets.
● Collaborate with Product & Design to expose explainer APIs ("Why Genie recommended this").
● Own real-time data pipelines (Kinesis / Kafka) feeding the graph with practice logs and class interactions.
● A/B test algorithms; lift practice minutes by ≥ 15 %, track WAU/MAU and retention impact.
Success Looks Like:
● Recommender boosts average practice minutes +15 % within 7 days of exposure.
● Gen-2 streak engine live by 15 Oct 2025 with
● Learner Graph uptime 99.9 %,