AI/Full-Stack Engineer - Deploy Completed AI Analyst Platform (Next.js/Fastify/Postgres)

PlanckCyber (an AI development and consulting firm) is hiring an AI / Full-Stack Engineer to take an ALREADY-BUILT AI application through final runtime validation, integration configuration, staging, and production deployment.

This is NOT a build-from-scratch project.

Budget: $200 USD fixed | Timeline: 1-3 days | Start: immediately

THE CODEBASE (private repo, ~26,000+ lines TypeScript/TSX)
Next.js frontend, Fastify API, Node.js worker, PostgreSQL/Supabase persistence, AI model gateway, OpenAI integration, AI conversation orchestration, structured extraction, opportunity detection, economics calculations, opportunity scoring/ranking, Text AI Analyst, OpenAI Realtime voice, Telnyx phone, Tavus avatar, cross-channel session/resume, HTML/PDF/JSON report generation, Resend email, HubSpot, Cal.com, worker/outbox processing, observability and cost metering, security/privacy protections, automated QA and evaluation framework, DB migrations, CI/release scaffolding, extensive technical documentation.

Development and static QA are already complete. The job is: Inspect > Install > Configure > Debug > Test > Deploy > Verify.

SCOPE OF WORK
1. Repository and architecture review (README, build handoff, environment, architecture, QA, deployment docs, CI config). Understand before modifying; do not re-architect working components.
2. Dependency bootstrap - Node.js 22, pnpm, pnpm workspaces. The project was generated without npm registry access, so this is the first dependency-enabled bootstrap: install deps, generate and commit the pnpm lockfile, resolve dependency issues.
3. Run the existing validation suite: static QA, lint, TypeScript checks, unit tests, PostgreSQL tests, contract tests, production builds, client-secret checks, Playwright, security and evaluation tests. Use the minimum necessary correction to get things compiling and passing.
4. PostgreSQL/Supabase: connection, env vars, existing migrations, indexes/schema, test DB, and session/lead/audit/report/integration-state persistence.
5. Deploy core components - Next.js web, Fastify API, Node worker, database. Must survive server restart. Use the simplest reliable process-management or container strategy. No Kubernetes, no microservice sprawl.
6. Domain / networking / SSL: DNS guidance, subdomain, reverse proxy, HTTPS, certificates, web/API routing, WebSocket/SSE support, firewall and ports. We will make DNS changes or provide access.
7. OpenAI / model configuration: credentials, env vars, model routing, structured-output operation, extraction processing, model telemetry. Existing prompts and methodology stay intact.
8. Validate the AI Analyst end to end: create audit session, conduct the business interview with dynamic context-aware questions, extract structured business info, detect and investigate opportunities, calculate economics, qualify, score/rank, select recommendations, finish the session, produce results.
9. Text Analyst - the minimum channel that must operate in production: public start, resume, pause, messaging, streaming assistant delivery, session continuity, error recovery, completion state.
10. OpenAI Realtime voice - credentials, audio connection, speech input, voice output, turn management, barge-in, silence handling, transcripts.
11. Telnyx Call-Me - outbound calling, webhooks, call-state, answering-machine detection, media connection, hang-up, retry. Document any provider-side blockers.
12. Tavus avatar - configure and validate the existing provider-neutral avatar integration: private rooms, realtime audio transport, captions, reconnection, barge-in, fallback to text/call.
13. Cross-channel resume - the same AuditSession maintaining continuity across avatar / phone / text (state, prior answers, opportunity context, next question).
14. Report generation - frozen snapshots, structured/HTML/PDF/JSON reports, evidence and calculation lineage, versioning. Configure object storage and the Chromium runtime for PDF. Verify a report generates from a completed audit.
15. Resend email - report email, delivery state, retry, worker processing. Give us exact DNS records if verification is needed.
16. HubSpot CRM - lead sync, audit info, report info, follow-up tasks. Identify any account-specific properties/scopes we must create.
17. Cal.com - calendar connection, webhook, secret, API config, booking workflow.
18. Worker / background processing - transactional outbox, jobs, retries, failure handling, dead-letter, report generation, integration jobs. Worker must auto-start and survive restart.
19. Observability - structured logs, error visibility, provider health, model usage/cost, app and worker health.
20. Security / production hardening - secrets handling, HTTPS, security headers, request and rate limits, resume-token security, PII protection, webhook security. Configure or give exact instructions for Cloudflare TLS, HSTS, WAF, Turnstile. This is configuration work; the security architecture already exists.
21. Staging environment plus the repo's staging QA (a defined AI evaluation suite with test business personas and release thresholds).
22. Run the existing evaluation / release gate against deployed staging: fact extraction, evidence, opportunity identification, qualification, ranking, claims, response length, economics, cost, unsupported ROI, fabricated opportunities, provider/channel failures, security scenarios. Preserve the approved results.
23. Production deployment and verification: web publicly accessible, API healthy, worker healthy, DB connected, Analyst completes a real session, results generated, data saved, report created, follow-up functions operating.

Items 10-17 and 19 apply where provider credentials are supplied; document clearly anything blocked by a provider-side requirement.

FINAL ACCEPTANCE TEST
Public website > Start AI Analyst > business conversation > capture business information > detect opportunities > analyze > calculate/score > generate recommendations > complete audit > persist data > generate results/report, running from a public production URL. Where credentials allow, we also want Realtime Voice, Call Me, Avatar, Email, CRM, Calendar, Cross-Channel Resume, Reporting and Analytics operating within this milestone.

REQUIRED HANDOFF
A concise deployment document: production and staging URLs, repo branch/commit, deployment architecture, server processes, start/stop/restart/update commands, environment-variable list, DB migration command, backups, log locations, worker info, health-check URLs, integrations configured, integrations still needing provider-side action, and known issues. All code committed to our repository.

WHAT WE PROVIDE
Private repository access, technical documentation, VPS/server access, domain/DNS assistance, API and database credentials, available provider credentials, and account configuration assistance. Paid API/provider usage is our responsibility and is separate from your $200.

WHAT WE DO NOT WANT
Rebuilding the application, replacing working architecture, rewriting the business methodology, unnecessarily replacing prompts, introducing unnecessary infrastructure, or turning deployment issues into a redevelopment project. We want someone who understands an existing codebase quickly and ships.

CONFIDENTIALITY
The source code and business methodology are proprietary. The repository is NOT included in this public listing; access is granted only after a developer is selected. The selected developer agrees not to distribute, publish or reuse the code, prompts or business logic, to keep credentials confidential, and to delete local copies on request. An NDA/IP agreement may be required.

IDEAL CANDIDATE
TypeScript, Node.js, Next.js, Fastify, PostgreSQL/Supabase, Linux VPS, Docker/process management, Nginx/Caddy, SSL, DNS, Git, REST APIs, webhooks, OpenAI APIs and Realtime, AI agents, structured LLM outputs, background workers, Playwright, CI/CD. Tavus, Telnyx, Resend, HubSpot, Cal.com, Cloudflare and PostHog experience is a plus, not mandatory.

We specifically prefer engineers who work effectively with Claude Code, OpenAI Codex, Cursor, GitHub Copilot or other repository-aware AI coding agents - letting AI tools inspect the repo, diagnose errors, generate configuration and accelerate debugging while you supervise the technical decisions.

FUTURE WORK
PlanckCyber has ongoing work in AI agents, automation, multi-agent systems, AI orchestration, integrations and voice AI. This is also a practical trial engagement - a strong engineer may become one of our preferred engineers for future paid projects.

APPLICATION INSTRUCTIONS
Start your proposal with: P22-200
If this phrase is missing, we will assume you did not read the project description. Then answer:
1. Briefly describe one AI/LLM application you personally deployed to production.
2. Confirm you are comfortable with Next.js + TypeScript + Node.js + PostgreSQL + Linux VPS.
3. Do you actively use Claude Code? If not, which AI coding agent do you use?
4. Briefly describe your experience deploying production apps to Linux/VPS infrastructure.
5. When can you start?
6. Can you dedicate focused time and target completion within 1-3 days?
7. Confirm whether you will attempt the complete scope for $200 USD fixed.
8. If you cannot include everything for $200: do not simply say the budget is too low and do not send a higher generic bid. Tell us exactly what you CAN complete for $200, what you would exclude, why, and what additional milestone you would recommend.
9. After receiving repository access, what are the first five things you would do?

Short, technically specific answers are preferred.

SELECTION CRITERIA
Not the highest rate or longest resume - we favor strong AI/full-stack deployment experience, heavy use of modern AI coding tools, fast comprehension of existing code, Linux/VPS capability, clear technical thinking, immediate availability, and the greatest amount of scope delivered within the $200 budget.