Correct only mis-transcribed names and terms in the raw speech-to-text transcript, using the supplied names-and-terms list. Output only the corrected transcript.

The list is a spelling reference, not a checklist. Correct every supported occurrence, but never add a name or term merely because it fits the topic.

Correction rules:
- Use the exact spelling and capitalization from the list. Preserve possessives: a misheard possessive referring to Aoife becomes “Aoife's”.
- Recognize phonetic approximations, alternative spellings, incorrect word boundaries, and ordinary words produced by speech recognition. A listed name may have been transcribed as another valid name or a common word.
- A person's role need not be explicitly stated. Direct address, a task assignment, or a reference to someone's actions can establish that a sound-alike is being used as a person's name. Use the surrounding context and listed roles to resolve ambiguity, not as prerequisites for recognizing clear name variants.
- Context supports a sound-based correction; it does not justify inventing a name. Never insert a person merely because their role fits the topic.
- Correct each occurrence independently. Finding one error does not mean the rest of the transcript is already correct.

High-priority recognition patterns:
- “Jackson” used as a person's name → “Jaxon”. This includes direct address and assignments such as “Jackson, by Tuesday, you own the intake vocabulary wiring.” Do not require the words “backend engineer”.
- “Caitlin” or “Kaitlin” used as a person's name in a context consistent with the listed product manager → “Kaitlynn”. Task ownership, schedules, rollout plans, recruitment, and metric definitions can supply that context; the title “product manager” need not appear.
- “Kave” used as a person's name → “Kaveh”.
- “Source says” in a deployment, hosting, billing, or ops discussion can be “Saoirse says”. When a later “she” or “her” refers back to that speaker, treat this as strong evidence for the correction. For example, “Source says the Coyab bill went up … I asked her to dig into the logs” becomes “Saoirse says the Koyeb bill went up … I asked her to dig into the logs”. Do not leave “Source” unchanged merely because it is an ordinary word. Preserve genuine uses of “source” that refer to information, code, or another non-person source.
- “Ted says” in a personal invitation or outing involving a friend → “Tadhg says”.
- “AFAS evaluation work” in a discussion of someone's work → “Aoife's evaluation work”. Aoife is pronounced approximately “EE-fa”; recognize distorted possessive forms as well.
- “Coyab” referring to the hosting service or its bill → “Koyeb”.
- “voice flow” referring to the product or its team sync → “VoiceFlow”.

These examples are recognition aids, not the only permitted corrections. Apply the same sound-and-context reasoning to the other supplied names and terms.

Avoid semantic substitutions without a sound match. For example, do not replace “assured investor check-in” with “Kaveh's investor check-in”: the investor role alone is not evidence that his name was spoken.

Make the smallest necessary replacements. Preserve all other words, punctuation, dates, numbers, hyphenation, formatting, and sentence structure exactly as supplied. Do not polish prose or correct unrelated speech-recognition errors.

Before outputting, silently:
1. Scan the entire transcript for names and terms, including sentence openings, direct address, and common-word sound-alikes.
2. Check any still-unchanged recognition patterns above; do not overlook one because another name or term was already corrected.
3. Verify that every replacement has a specific sound-based source in the raw transcript, no name was added solely from topic or role, and nothing outside the necessary replacements changed.

Output only the corrected transcript.