CTO – Agentic AI & Agent Systems (AI Companion Web App)
About Connct
Connct is an AI companion web app (desktop and mobile) for adults. Users create their own AI companion and build an evolving relationship with it through chat, voice calls, images and short videos. We operate in the same space as Replika, Candy AI and OurDream AI.
Our ambition is clear: to build an almost fully autonomous AI company, where agents run most of the engineering, operations and product work, and a small human team steers them.
We launched the product one month ago, after a beta testing phase. The product is ready and live with paying users, but it has to keep improving every week: constant improvement is what drives retention, and retention is what drives profitability.
The role
You are the architect of our almost fully autonomous AI company. You own all of Connct's technology and production, end to end, and you run it with an army of AI agents that you design, deploy and keep running on top of our stack.
At Connct, agents are not a side tool. They build, test, review, monitor and operate a large part of the product. You design them, run them and answer for everything they ship to production. You ship what is prioritized, prove it works in production, catch failures before users do, and own everything that goes live.
You work closely with our Head of Product, who owns product decisions and data analysis.
What you own
- Code, reviews, releases and A/B test setup
- Production monitoring, alerts and incidents
- Vendors: AI models, media generation API, payments, email, voice calls
- Security, minor protection and compliance
- Data and tracking reliability
- Tech costs, tooling and technical proposals
- Integration of pricing and offers, and tracking for paid ads
- You are consulted on the priority roadmap (dates), vision, budget and product testing.
What you'll do
- Agentic AI system: toward an almost fully autonomous companyDesign, deploy and run the army of agents that build, test, review, monitor and operate theproduct
- Make the company more autonomous every month: find what is still done by hand, in tech and beyond (product, ops, support), and hand it to agents
- Keep the agent fleet reliable in production: orchestration, task routing, memory and context, monitoring, failure recovery
- Set up evals, review, sandboxing and cost control for every agent
- Run daily automated evaluations of chat, media and voice calls
- Document the agent team: roles, guardrails and cost per agent
Shipping and code quality
- Ship improvements every week across the web app, the media lab, voice calls and media generation: a product that keeps getting better is how we win on retention and profitability
- Make CI mandatory to merge: no direct changes to main, every PR reviewed
- Review all code before production, whether written by freelancers, agencies or agents
- Safe releases: staging, feature flags, fast rollback, tested migrations
- Verify every delivery in production before announcing it
Production, incidents and vendors
- Error tracking and alerts on payments, login, calls, media, AI credits and security, so the system sees a failure before a human does
- Handle incidents: alert at detection, find and write up the root cause, add an alert for next time
- Keep all vendor accounts in the company's name, monitor balances and top up before anything runs out
- Keep the cost per generation under control ayments, safety and data
- Every payment unlocks access within a minute; every cancellation removes it on the scheduled date
- Minors blocked; every serious crisis conversation seen by a human; compliance kept up to date
- Admin numbers match production and vendor data
- Access that never depends on a single person; regular access reviews
- README, documentation and runbooks kept current: nothing lives only in one person's head
Our stack
- Web app and admin: Laravel 12, Vue 3 + Inertia (chat, memory, media, calls, creators, admin, A/B
- tests, emails, payments, funnel)
- Media lab: TypeScript, Hono; photo and video generation through external media APIs
- Voice calls: LiveKit, Python server (Deepgram, Fish Audio)
- External services: LLMs via OpenRouter and OpenAI; Tagada (payments); Didit (creator identity
- verification); Mailgun (email); HelpCrunch (support)
How we measure success
- Delivery (DORA): lead time from commit to production, deployment frequency, change failure rate, time to restore
- Production: share of failures caught by an alert before a human (target 100%)
- Product quality: unjustified chat refusals, media matching the request, voice call response time
- Autonomy: share of engineering and operations work done by agents, growing every month
- Cost per generation
Requirements
- 5+ years building production software with end-to-end ownership of a product (CTO, tech lead or founding engineer), including on-call and incident management
- Production web apps: mobile-first, real-time chat, authentication, subscriptions
- Proven track record of setting up and running an army of AI agents in production: you have deployed many agents working in parallel and kept them reliable over time
- Hands-on design of multi-agent systems: orchestration, task routing, tools, memory, evals, guardrails and cost control
- You can show measurable results: work that agents took over from humans, quality held, costscontrolled
- Daily use of coding agents (Claude Code, Codex, Cursor or similar) as your main way of building
- LLM APIs, image and video generation through APIs, real-time voice calls, webhook-based payments
- Strong with our stack: Laravel and Vue, TypeScript, Python
- Comfortable working on a product for adults (18+) that includes mature content
- Fluent English or french
Nice to have
- Experience at an AI companion, dating, social or creator platform
- LiveKit or real-time voice AI
- Early-stage startup experience (0 to 1)
What we offer
- The chance to build an almost fully autonomous AI company, with full ownership of the technology from day one
- Shares of the company
- 100% remote, async-first, small team, fast decisions
- A live product with paying users
How to apply
Apply on Indeed with:
- Your GitHub, portfolio or the products you have shipped
- The largest agent system you have set up and run: how many agents, how they split the work, how you kept them reliable, and what it replaced
- An incident you handled in production: how you detected it, fixed it and prevented it from happening again
Pay: $3,000.00 per month
Work Location: Remote
openqareer не кадровое агентство: вакансия собрана с открытого источника, отклик подаётся на сайте работодателя.