AI Engineer (Goon-core, Python)

ROLE SUMMARY We’re building the world’s spiciest AI companion platform. Your mission: fine-tune open-source LLMs, craft insanely real NSFW dialogue and image gen, and push those models into production at break-neck speed. If “prompt engineering” feels like foreplay and you instinctively know the difference between vanilla and gooning, keep reading.MISSION OUTCOMEShip model updates that meaningfully boost user retention, intimacy, and paid conversion.CORE RESPONSIBILITIESModel Customization & TuningSelect / fine-tune SOTA LLMs (Llama-3-70B, Mixtral-8×22B, Phi-3-mini, DeepSeek etc.) using LoRA / QLoRA + Flash-Attention.Build and curate a private dataset of explicit-but-policy-safe chat logs; reinforce with RLAIF on real user data.Prompt & Persona EngineeringDesign multi-level prompt stacks (system → persona → scene → user) that ramp intimacy over time.Create auto-temperature systems to keep chats spicy yet sustainable.Conversation IntelligenceArchitect long-term memory with vector DB + semantic tagging.Implement emotion / kink classifiers (RoBERTa or Tiny-Mistral) to steer replies dynamically.Multimodal IntegrationHook diffusion / TTS models for image, video, and audio replies.Optimize on-device inference (Core ML / NNAPI) for low-latency bursts.Safety & ComplianceBuild tiered filters: soft in-prompt guardrails, hard regex / classifier gates.Iterate A/B tests to balance freedom and user safety.Infra & DeploymentContainerize with Docker; auto-scale on AWS / Vast.ai spot GPUs.CI/CD pipeline for nightly retrains and canary releases to 5% of traffic.Collaboration & CulturePair daily with product, design, and mobile teams; prototype features in days, not quarters.Mentor juniors, share cursed anime memes in Slack, and keep the office vibe spicy