Headquarters: Tampa, Florida

URL:https://koast.ai

Head of Product | Contract, 20 hrs/week to start, growing into full-time | Remote (US overlapping hours) | Reports to the CEO

The short version

Koast is the AI operating layer for teams that run Meta ads at volume. Agencies, in-house media buying teams, and enterprise buyers managing 10+ ad accounts and launching 50+ ads a week use Koast to launch faster, automate optimizations off real attribution data, and see every account from one hub. The Koast Agent is the next layer: an AI that reads the account, spots what matters, and acts inside guardrails.

We are a small team shipping fast. The product is live, customers are paying, a Whop partnership just onboarded a wave of new users, and the Agent build is underway. What we do not have is one person who owns the product end to end: what gets built, why, in what order, and whether it actually landed.

That is this role. It starts at 20 hours a week on contract. If you are the right person, it becomes the full-time product seat at Koast.

What you will own

Discovery that finds the real problem. You will sit in on enterprise sales calls and onboarding sessions, run discovery with agencies and media buying teams, and figure out what they are actually trying to get done versus what they say they want. Every feature request gets mapped to a job: launching smoother, getting their team in, getting tasks done, keeping users active and growing spend inside Koast. If it does not serve a job, it does not get built.

Specs that engineering can execute without a meeting. We run a lean engineering team plus an AI-assisted dev pipeline. That only works when the ticket is unambiguous. You will write PRDs and Shortcut tickets that define the behavior, the edge cases, the data dependencies (Meta API limits, Hyros and Cometly attribution feeds, Whop account provisioning), and the acceptance criteria. If an engineer or an agent has to guess, the spec was not done.

The queue, and what is missing from it. You own the Up Next column. You will keep engineering fed so they are never blocked, and you will keep a standing view of what is discussed in Slack, support, and QA that does not yet have a ticket. Gaps are your problem before they become customer problems.

Evidence over opinion. PostHog, Intercom, session replays, rage clicks, support threads. You will run the user-behavior audit, know where people drop off in launching, and bring numbers to every prioritization argument, including the ones with the CEO.

Delivery from spec to shipped. Governance, milestones, release notes, and the customer conversation after it ships. When something breaks or slips, you are the one who says so early, not the one who explains it after.

Technical judgement in a non-deterministic product. The Koast Agent makes decisions off data that is noisy, seasonal, and shaped by Meta's learning phase. You will make product calls that respect that: what the Agent is allowed to do, what it must never do, how we measure whether an automated action was correct. Guardrails are infrastructure here, not UI. You need to understand why.

What you will do week to week

Run discovery sessions with prospects and customers and turn them into scoped, execution-ready PRDs.
Surface technical risk early: undocumented API behavior, missing data, an assumption that will not survive a real ad account.
Write and groom Shortcut tickets with enough detail for an engineer or Claude Code to ship without a follow-up question.
Partner with engineering on feasibility, sequencing, and cutting scope to what ships this week.
Own success metrics per feature and report on them after launch, not before.
Be the customer's point of contact through onboarding and rollout for enterprise accounts.
Write the updates: one version for the CEO, one for the customer, one for the engineer. Each one short.
Say "we should not build this" when the data says so.

What sets you apart

You have run Meta ads or sat close enough to a media buyer to know what a 2am budget change feels like. You know what Advantage+ does to a structure, why learning phase resets matter, and why "just pause the losers" is a bad automation.
You have shipped B2B software end to end in a small team. Not managed it from a distance. Shipped it.
You have worked on or with AI or ML products where the output is not deterministic and you had to define what "correct" means.
Senior engineers take you seriously b

Head of Product @ Koast.ai at Koast.ai