Agentic Experience Engineer

About our vision
The future of software is not software. It's a constellation of agents that know you, act for you, and answer only to you — living on device with the whole family’s safety in mind, governed by a constitution of care rather than a business model. It's agents that delight those who use them, and earn for those who shape them. All without ads, attention harvesting, or data leaving your device unless you say so.
Our AI framework isn't built on any one LLM. It's a single agentic harness for any model we choose to put under the hood. The constitutional governance, the memory architecture, the safeguarding layer — that's the IP. The agents that bring it alive across multiple high-value sectors? They’re yours.
About the opportunity
Maloka Labs is a British AI startup that has spent the last year heads down, quietly working out the shape that infrastructure for trusted consumer AI should take. Now we're building it.
We need a new kind of developer to own the layer where our AI agents meet the real world — proving they behave as promised, keeping personal data private on device, and orchestrating emotionally intelligent AI guides that serve influencer partners in high-value consumer sectors, and their communities. We call this Agentic Experience Engineering, or AXE for short.
You'll be joining a global, 12-strong team of storytellers, education and consumer-sector professionals, AI and backend engineers, platform and blockchain architects, data specialists and financial innovators. And a hippie or two (including the founder, but don't tell him we said that).
You'll work in close partnership with our Chief Hippie, umm, Imagineer, plus our Head of AI Engineering, and our CTO — building alongside frontier models while governing quality and constitutional compliance. In short, you will own how our family of agents look, sound and behave.
What you'd own
• Agent evaluation and correction: The harness that tests every agent against our constitutional framework before release, and the human-in-the-loop cycle that catches drift and feeds it back into refinement. The seat's central function — and the source of the workflows we demonstrate to investors and institutional partners.
• Behaviour implementation: Turning constitutional specifications into testable conduct — what an agent may say, and when it defers to a human.
• Voice interface, on-device inference and serving: Natural voice interaction within latency and quality thresholds, and governed agent behaviour running privately on the user's device. Built in collaboration with our Head of AI Engineering and CTO.
• Influencer partner targeting and onboarding: A pipeline that qualifies partners for our influencer-related services, built once for rapid adaptation within multiple family-facing verticals.
• Agent-driven distribution: Oversight of our unique agentic sampling programme — every micro-experience on external platforms nourishing rather than manipulating, and inviting users into the ecosystem.
• Local-first data and privacy: On-device memory, P2P sync, and secure local persistence — the agent layer must embody our privacy promise, not just sit on top of it.
• All you can eat: The agentic layer is the floor, not the ceiling. Bring your own appetite for what this role can become.
Ideal Haves
• Multi-agent orchestration — agents that coordinate, hand off, and hold coherent state.
• RAG pipelines that ground responses in domain knowledge, not free association.
• AI safety and evaluation in production — guardrails, red-teaming, or content-safety systems for real users.
• On-device or edge ML experience, or the appetite to master it fast.
• Production engineering in Python and/or TypeScript — shipped services, not just notebooks.
• A feel for what makes AI interactions feel human — flow, tone, memory, and graceful failure.
• You translate fluently between engineers, creatives and partners — complex ideas into clear, buildable specifications.
• Constraints energise you — guardrails and privacy-first architecture feel like creative challenges.
Stretch and bonus
• DPO, RLHF, or preference-based fine-tuning for behavioural alignment.
• Model quantisation, or voice pipelines that hold persona consistency across voice and text.
• Built or contributed to an agentic framework (LangGraph, CrewAI, AutoGen, or your own).
• Agent governance layers, or published safety research.
• Work with family-oriented products.
• Privacy-enhancing techniques, or EU AI Act, GDPR and COPPA compliance.
• TON or Telegram Mini Apps.
About our offer
• Package: Compensation befitting your experience, plus founding team equity share (vested over three years, one year cliff), with potentially life-changing upside for exceptional contribution. Generous signing bonus and benefits package on close of our raise.
• Commitment: Full-time on close of our raise, possibility of part-time agreement before.
• Domain: You own the agentic experience layer, end to end.
• Founder access: Weekly vision sessions with our Chief Imagineer, plus daily async collaboration.
• Peers, not managers: You'll work alongside Chiefs and Heads, not under them.
• Location: Remote, async-by-default across timezones.
• Pathway: This role has a documented pathway to Head of Agentic Engineering and beyond, with a possible pathway to CPO for the right individual.
This is not just another job building just another AI app
• You're not building a chatbot — you're engineering the behaviour layer of a constitutionally governed ecosystem, and the proof it keeps its promises.
• You're not optimising engagement metrics — you're architecting agents that work better the less they phone home.
• You're not staring at a blank repo on day one — you're inheriting a fully specified agentic architecture, a constitutional governance framework, and a team that spent the last year in R&D stealth, architecting and blueprinting the plans from which we are now building.
• The agents you build will serve our influencer partners and the millions of people they support — including teenagers through our nonprofit. Setting the highest safety benchmarks matters.
One more thing
Like you, we're hungry too; hungry to get to market and up our raise. In this launch sprint everyone has their sleeves up, everyone builds — you'll be hands-on in the codebase, serving as human-in-the-loop with our AI co-builders (currently Opus 5.5 and Sonnet 5.5), not directing from a whiteboard. The team you'll shape comes later; first, we ship v1 together. If you're too proud to push code, then swipe left.
Apply
Don't just click Apply — message us: "AXE" + your proudest agentic build and what made it behave, or why its juice is ready to squeeze into our cocktail. Preferably with a link.
We'll continue reviewing messages after this ad closes.
No recruiters. No cover letters. Just proof you know how to ship.
Carpe diem. Legacy counts.

Work arrangement
Yes