B Babelio Playbook 11 / 24
Refresh: quarterly EN · RU
Operating System · Artifact 11

Hiring Plan

Revenue-keyed hiring sequence + five Topgrading scorecards for the first hires. One hire is load-bearing: a Rust + CoreAudio/WASAPI audio engineer the 10-week MVP and the entire moat depend on — and who is not yet committed. Hire on Monday using only this doc.

First hire
Audioeng
Loaded / mo
$14K
Scorecards
5
Sequencing
MRRnot dates
▲
Hire #1 is the company's single point of failure

The Rust + CoreAudio/WASAPI audio engineer builds the per-process audio capture (macOS Tap API + Windows WASAPI loopback) that browser extensions structurally cannot do — the reach head-start the whole product rests on. The 10-week MVP timeline assumes day-1 access to this person (product.md §3). They are not yet committed: REVIEW.md flags founder-market-fit on OS-internals as a P0 gap (VC Team 0/50, "no FMF = no meeting").

At ~$14K/mo fully loaded this one person is ~70% of R&D burn through Q1–Q3. Until this seat is filled, no other hire matters and no timeline is real. Treat closing this hire (or naming a committed co-founder/contractor) as the first deliverable of the whole playbook.

12-month sequence · keyed to revenue, not dates

Each hire fires on an MRR milestone, not a calendar date. The triggers below map 1:1 to the headcount roll-forward in 10-financial-model.html. This is a community/content-led consumer motion (06-gtm-motion.html) — engineering is hired ahead of revenue; everything else trails it.

Role Revenue trigger Base mo. Loaded $/mo Equity
Audio / OS-internals eng load-bearingPre-MVP · Day 0M1$14,0001.0–3.0%
2nd engineer (full-stack)$10K MRRM10$12,0000.5–1.0%
Growth / community lead$25K MRRM14$8,0000.3–0.7%
3rd engineer (Windows/edge)$40K MRRM19$11,0000.3–0.6%
Support / CS lead (part-time)$40K MRRM20$4,000—
Bar for good: a non-recruiter founder can run any of these five searches start-to-offer using only this page — JD, scorecard, interview loop, and reference script are all here.

Scorecard 1 · Audio / OS-internals engineer

Audio / OS-internals engineer

Hire #1 · load-bearing · Rust + CoreAudio/WASAPI
Day 0

Mission: ship low-latency per-process audio capture + auto-mute on macOS and Windows so Babelio can translate native desktop apps browser extensions can't reach.

Outcomes (measurable)
  • O1macOS CoreAudio per-process Tap capturing ≥3 real native clients (Webex, an LMS player, VLC) by Week 4.
  • O2Windows WASAPI per-session loopback + per-session duck on ≥3 native clients by Week 6.
  • O3End-to-end glass-to-glass p95 < 700ms on the latency-budget harness by Week 8.
  • O4Per-app capture + eval telemetry logging from session 1 (the moat flywheel) — live by Week 6.
  • O5Signed/notarized macOS .dmg + signed Windows .msi with auto-updater shipped by Week 10 (MVP done).
Competencies
Rust (production) CoreAudio Tap API WASAPI / audio sessions Real-time / low-latency DSP Tauri / FFI Code-signing / notarization Ships solo, no hand-holding Pragmatism over purity
Salary$140–180K + loaded $14K/mo
Equity1.0–3.0% (co-founder-grade if pre-raise)
Interview bank · 10 questions
  1. Walk me through capturing one process's audio on macOS 14.4+ without a kernel extension.
  2. How do WASAPI loopback and per-session ducking differ from system-output capture?
  3. Where would you spend a 700ms glass-to-glass budget across capture, VAD, STT, MT, TTS?
  4. Tell me about a real-time audio bug you root-caused. What was the actual cause?
  5. Rust↔Python sidecar over local UDS/ws: where do you put the buffer and why?
  6. macOS notarization keeps failing on a Tauri build — how do you debug it?
  7. A native client routes audio through a path your tap misses. How do you find it?
  8. What would you instrument from capture session 1 so the eval flywheel compounds?
  9. You have 10 weeks and one shippable OS. Which do you ship first and why?
  10. Why this problem, for a year, when bigger comp exists elsewhere?

Scorecard 2 · Full-stack engineer

Full-stack engineer (#2)

Pipeline + web/billing · Python + Next.js
$10K MRR

Mission: own the Python voice pipeline + Next.js account/billing layer so the audio engineer stays on native capture and the product can scale paid users.

Outcomes (measurable)
  • O1Metered usage_ledger + billing reconciliation live within 6 weeks; COGS reconciles to within 5% of provider invoices.
  • O2Per-user budget cap → auto-degrade dub→subtitle shipped; no user exceeds plan COGS silently.
  • O3LiteLLM fallback path: a provider 5xx mid-session never drops the dub (verified in chaos test) by month 2.
  • O4Promptfoo + latency-budget evals gating every deploy; zero un-gated prompt/model swaps reach prod.
  • O5Self-serve signup→paid funnel instrumented; activation events flow to the KPI dashboard.
Competencies
Python asyncio / FastAPI Next.js / Vercel Postgres / billing LLM/STT/TTS provider APIs Eval/CI (Promptfoo) Observability (OTel) Generalist range Ownership without specs
Salary$120–150K / loaded $12K/mo
Equity0.5–1.0%
Interview bank · 10 questions
  1. Design the usage_ledger row so COGS reconciles per billable second. What columns?
  2. A provider returns 5xx mid-session. How does your fallback keep the dub alive?
  3. When does a user hit their budget cap, and what degrades first?
  4. How would you stop an un-evaluated prompt change from reaching prod?
  5. Walk me through a metered-billing edge case you got wrong once. What broke?
  6. Where do you draw the line between your work and the audio engineer's?
  7. A signup→paid funnel drops 60% at install. How do you isolate the cause?
  8. Persistent provider websockets vs per-utterance connect — tradeoffs?
  9. You're the only one who can fix a prod billing bug at 11pm. What do you do?
  10. What part of this stack would you refuse to over-engineer, and why?

Scorecard 3 · Growth / community lead

Growth / community lead

Community-led acquisition · not a sales hire
$25K MRR

Mission: turn immersion-learner communities into a repeatable, sub-$35-CAC acquisition engine and own the share-loop content from Session Review.

Outcomes (measurable)
  • O1Blended CAC held ≤ $35 while net-new paid grows ≥ 15% MoM for 2 consecutive quarters.
  • O23 named immersion-learner communities converting at the projected rate, documented as a repeatable playbook.
  • O3Bilingual-clip share loop live; ≥ 10% of new signups attributable to user-shared Session Review clips by month 4.
  • O4Content/SEO engine producing 4 ranking assets/mo that compound organic signups.
  • O5Lifecycle activation lift: free→paid conversion improved toward 10% via onboarding experiments.
Competencies
Community building Content / SEO Lifecycle / activation Analytics-literate (CAC/LTV) Experiment design Speaks the learner's language Scrappy / hands-on Brand voice / writing
Salary$80–110K / loaded $8K/mo
Equity0.3–0.7%
Interview bank · 10 questions
  1. Name 3 immersion-learner communities and how you'd earn trust in each.
  2. Walk me through a community-led channel you grew. What was the CAC?
  3. A single-player product has a weak viral loop. How do you build one anyway?
  4. How would you make a bilingual Session Review clip worth sharing?
  5. CAC creeps from $35 to $60. What's your first three diagnostics?
  6. Free→paid is stuck at 6%. What onboarding experiment do you run first?
  7. Which metric would you refuse to optimize because it's vanity?
  8. Show me a piece of content you wrote that converted. Why did it work?
  9. How do you avoid burning a community's goodwill while acquiring from it?
  10. We can't hire a sales team yet. How does growth happen without one?

Scorecard 4 · Engineer #3 (Windows/edge)

Engineer #3 — Windows / edge

Windows depth + on-device STT pilot
$40K MRR

Mission: harden Windows capture across the long tail of native clients and pilot on-device STT to kill the dominant $0.55/hr STT COGS leg.

Outcomes (measurable)
  • O1Windows capture verified across 10+ native clients (Teams, games, regional tools); < 2% capture-failure rate.
  • O2On-device Whisper-turbo + Silero STT path live where hardware allows; cuts STT COGS ≥ 40% for those users.
  • O3Per-OS p95 latency parity maintained: Windows within 10% of macOS by quarter-end.
  • O4SmartScreen/antivirus install-friction reduced; verified-publisher path on signed .msi cuts first-run drop-off.
Competencies
Windows audio internals Rust / C++ On-device ML (ONNX/Whisper) Code-signing / SmartScreen Perf profiling Cross-OS parity mindset Long-tail debugging patience Cost-aware engineering
Salary$110–140K / loaded $11K/mo
Equity0.3–0.6%
Interview bank · 10 questions
  1. A game routes audio through an exclusive-mode device. How do you still capture it?
  2. Where would on-device Whisper-turbo actually beat cloud Deepgram on cost and when not?
  3. How do you keep Windows p95 within 10% of macOS on weaker hardware?
  4. SmartScreen flags our signed .msi anyway. What's your remediation sequence?
  5. Tell me about a long-tail platform bug only some users hit. How'd you reproduce it?
  6. When would you choose to NOT run STT on-device despite the COGS win?
  7. How do you profile a real-time audio path without distorting its timing?
  8. The macOS engineer made a design choice that hurts Windows. How do you resolve it?
  9. A capture failure happens for 2% of sessions silently. How do you surface and fix it?
  10. What's the cheapest experiment that would tell us if on-device STT is worth it?

Scorecard 5 · Support / CS lead

Support / CS lead (part-time)

Trust, install-friction, retention signal
$40K MRR

Mission: own the first human touch — install/permission friction and live-session issues — and turn support tickets into the early-warning churn signal.

Outcomes (measurable)
  • O1First-response < 4 business hours; install/permission tickets resolved < 24h at ≥ 90% CSAT.
  • O2Self-serve help center covers the top 20 install/permission issues; deflects ≥ 50% of tickets by month 3.
  • O3Weekly churn-signal digest to founder: top cancel reasons + at-risk accounts feed retention experiments.
  • O4Verified bug reports triaged into the eval gold set; recurring failure modes become labeled clips.
Competencies
Technical empathy Install/OS-permission savvy Clear writing Triage / prioritization Bilingual (EN/RU+) a plus Pattern-spotting from tickets Calm under angry users Self-directed
Salary$48–60K FTE-equiv / loaded $4K/mo PT
Equity— (options pool later)
Interview bank · 10 questions
  1. A user is scared by a Gatekeeper "unidentified developer" warning. Talk them through it.
  2. An angry user says the dub echoes badly. What do you ask, in what order?
  3. How would you turn a week of tickets into one churn-signal digest for the founder?
  4. Which 20 articles would you write first to deflect the most tickets?
  5. A bug needs the engineer but they're heads-down. How do you escalate well?
  6. Tell me about a time you spotted a product problem from support patterns.
  7. A user reports a translation error. How do you turn it into an eval example?
  8. You can't fix something today. How do you keep the user's trust anyway?
  9. How do you decide a ticket is a one-off vs a systemic issue?
  10. Privacy-anxious user asks what we do with their meeting audio. Your answer?

Reference-check script · 5 questions

Run on every finalist before an offer. Ask for references the candidate did not volunteer (past manager, peer). Listen for hesitation as hard as for answers.

  1. Context: "How did you work together, and for how long?" (verify the relationship the candidate described)
  2. Outcomes: "What did they actually ship, and what was hard about it?" (map to the scorecard outcomes above)
  3. Ranking: "Of everyone you've worked with in this role, what percentile were they?" (a vague top-10% with no example is a flag)
  4. Growth edge: "What would they need a stronger teammate around them for?" (the honest weakness)
  5. Rehire: "Would you hire them again, and for exactly what?" (silence before "yes" tells you more than the yes)
Decision rule: two independent references must each give a concrete example tying to a scorecard outcome. Generic praise ≠ a pass. For the audio engineer, at least one reference must speak to shipping real-time audio across an OS boundary, not just "good Rust dev".

Founder no-go list

Who NOT to hire — and when
  • ✕No VP of Marketing / sales team pre-$40K MRR. This is a community/content-led consumer motion (06-gtm-motion.html); a sales org pre-PMF burns the runway the audio engineer needs.
  • ✕No standalone designer before MVP ships. A thin Tauri web UI on a few screens (product.md §5) doesn't justify a full-time hire; contract for the install/trust flow instead.
  • ✕Don't hire a generalist "AI engineer" as hire #1 instead of the OS-internals specialist. The hard, rare, moat-defining work is per-process audio capture — not prompt/model wiring. Substituting here is the most expensive mistake on this page.
  • ✕No senior manager / "head of eng" before there are 3+ engineers. Layers before there's a team to manage is pure burn; the founder runs eng directly until hire #3.
  • ✕Don't hire to fix the founder-market-fit gap with a junior. The OS-internals seat needs a proven senior (or co-founder); a junior here means the 10-week MVP and moat both slip (REVIEW.md P0).
Cross-references
  • 10-financial-model.htmlHeadcount roll-forward — the revenue triggers and loaded costs above map 1:1.
  • product.md §310-week MVP scope + the load-bearing audio-engineer contingency.
  • tech.md §1, §8Stack the hires build on; STT-dominated COGS that hire #3's on-device pilot attacks.
  • 06-gtm-motion.htmlCommunity/content motion that defines the growth-lead role and the no-sales-team rule.
  • 16-risk-register.htmlFounder-market-fit / hire-#1 gap carried as a P0 risk with a tripwire.
Babelio · Hiring Plan · Playbook 11/24 Hire on Monday using only this doc · refresh quarterly