You will receive a raw speech-to-text transcript and a list of names and terms. Correct misrecognized occurrences of those names and terms to their exact listed spellings. Output only the corrected transcript.

Make targeted replacements, not a general rewrite:
- Preserve all other wording, punctuation, capitalization, repetitions, and formatting.
- Leave correctly written names and terms unchanged.
- Do not add information, complete sentences, or introduce a listed name or term merely because it fits the topic.
- The list is a correction vocabulary, not an exhaustive list of everything the speaker might mention. Unlisted names and technical expressions can be valid.

Identify corrections using pronunciation and local context, not spelling similarity alone. Speech recognition may turn an unfamiliar name into a familiar name, an ordinary word, or a shortened fragment. Check grammatical cues such as a person doing something, direct address, and possessive references to someone's notes or work. Use the supplied roles as supporting evidence.

Examples of plausible name corrections when context supports them:
- “Jackson” → “Jaxon”.
- “Merritt sent me her recipe” → “Mairead sent me her recipe”.
- “Pre's API notes” → “Priya's API notes”.
- “Sarah's rollout note” → “Saoirse's rollout note” when referring to the deployment/ops person's work.
- “search's less friendly replay” → “Saoirse's less friendly replay” when the phrase refers to a person's replay or test rather than to search functionality.

These are contextual corrections, not unconditional substitutions. Preserve genuine mentions of other people and ordinary uses of words such as “search.”

Be especially careful with short fragments and acronyms. Require recognizable phonetic evidence for the particular listed term; technical context alone is insufficient. For example, “loads the M file” in a discussion of environment configuration and API keys does not justify inserting “MLX.” Leave such text unchanged unless it clearly represents a listed term. Do not repair unrelated transcription errors.

Before outputting, check for overlooked names, including possessives and names transcribed as common words. Then verify that every replacement is supported by both the original sound pattern and the sentence context, and that nothing else changed.