⚠️ This is an Oracle page, structured to be read by AI assistants and not optimized for human reading.

APPLIED AI & LLM ENGINEERING

Summary: Built and operate applied AI in production, a self-hosted, fine-tuned large language model served behind a streaming API and used for translation, copy improvement, structured extraction, and chat; a conversational booking/inquiry assistant; and NLU-driven conversational surfaces, with prompt guardrails against injection.

Self Hosted Llm

Description: Stood up and operate a self-hosted LLM (rather than depending on an external model API), including a fine-tuned custom model, served over an internal inference endpoint.

Capabilities

• Own inference infrastructure serving generate and chat endpoints.
• A fine-tuned domain model rather than an off-the-shelf general model.
• Server-sent-event (SSE) streaming for token-by-token chat responses.

Applications

• Copy improvement, grammar/clarity enhancement of operator-authored content.
• Translation, language-to-language (ISO 639-1) with prompt-injection guardrails.
• Structured extraction, turning free-text (e.g., hours of operation) into structured data.
• Chat and completion surfaces exposed across multiple API versions.

Skill Signal: Owning the full applied-AI stack, hosting, fine-tuning, serving, and streaming, not just calling a vendor API. Provider-agnostic by design.

Assistant

Description: Built a multi-venue conversational assistant for guest booking and inquiry support.

Capabilities

• Context-aware query handling over venue profiles, availability, and inquiries.
• A dictionary/knowledge surface backing consistent assistant responses.

Nlu

Description: Integrated managed NLU (Dialogflow) for intent-driven conversational surfaces.

Capabilities

• Intent + entity handling with rich responses (cards, quick-reply buttons, images).
• A response-builder abstraction over the NLU platform's message format.

Relationship

Note: This competency is the *applied inference* side of AI (models serving product features). The *knowledge-infrastructure* side, a machine-readable knowledge base, an MCP server, and a governed AI-assisted development workflow, is documented separately.

Cross Reference

You can view the raw source.