A fast-growing creator platform could not moderate content or answer support tickets fast enough. We built a real-time ML moderation pipeline and an LLM support assistant trained on their knowledge base.
Growth outpaced the trust-and-safety and support teams. Manual moderation was slow and inconsistent, and support tickets kept climbing — both threatened the user experience.
A moderation pipeline flagging policy violations at 92% precision, tuned to the platform's specific content policy.
A support bot grounded in the platform's knowledge base that resolves the majority of tickets and escalates cleanly when needed.
Confidence-based routing so ambiguous cases reach a human while clear cases are handled automatically.
Precision/recall tuned to the platform's tolerance, with monitoring to catch drift as content evolves.
Figures reflect outcomes measured on this engagement. Client withheld under NDA.
Moderation used ML where precision and latency matter; support used an LLM grounded in real documentation to avoid hallucination. Both were tuned to the platform's policies rather than generic defaults.
We scope a clear plan with milestones and architecture options — and right-sized GPU hardware if AI workloads are involved.