Build autonomous systems with the people who ship them
Kenjin runs on a small, senior team working directly with the founders. AI workforces handle the volume; humans handle the judgement. Both open roles span client work and the development of our own products — several launching over the coming year. If that's how you'd want to work, read on.
Three things worth knowing before you apply
Orchestration, Always
All substantial work runs as a premium-model coordinator briefing an AI workforce. The coordinator owns design, contracts, verification and the hands-on pass. You'll spend more time designing, briefing and verifying than typing.
The Engineering Ethic
We build the simplest extensible thing. No hacks, no bandaids under schedule pressure. If a spec can't be met without violating that, you stop and report — and you're never penalised for stopping.
Verification Integrity
Green test suites are necessary, never sufficient. Nothing is "production-ready" without evidence in hand and a plain statement of what wasn't covered. Conflicts are raised as open rulings, never resolved silently.
Two roles, London (hybrid)
Senior Agentic AI Engineer
You'll build the agent teams, memory infrastructure and data pipelines behind Kenjin's client systems and products — on the Mastra/TypeScript stack, under our written engineering standard. Senior, hands-on, and trusted to act as coordinator on your own lanes: verifying facts, briefing AI workforce agents, running gates, and doing the hands-on verification yourself.
- Shipped multi-agent or agent-workflow systems that run unattended in production
- Strong TypeScript/Node.js, Postgres, vector/RAG patterns, LLM APIs and local inference
- Real applied ML on behavioural or content data, with proper evaluation
- Serious experience using agentic coding tools as a workforce — briefing, gating, verifying
Full role details
What you'll do
- Architect and build multi-agent systems on Mastra: roles, delegation, workflows, tool integrations, MCP servers, evaluation loops
- Implement three-tier memory (agent / team / institutional) on native memory types and vector store
- Build continuous aggregation pipelines from news, market, API, social and live sources into structured, queryable knowledge
- Engineer analytical AI: predictive models for content performance, audience identification, behavioural analytics
- Design human-in-the-loop governance: intervention points, audit trails, approval gates, fail-closed checks
- Operate as coordinator on your lanes and keep documentation current in the same change as the work
Bonus
- Multi-source, real-time data architectures (news, financial, social platform APIs, streaming telemetry, sports data)
- Data science depth — feature engineering, time-series, experimentation; Python alongside TypeScript
- Media, streaming, broadcast, stock/DAM, archive, sports or talent-industry experience
- Multi-modal pipelines, IPTC/XMP metadata standards, C2PA
- Electron/Tauri desktop packaging, code signing, auto-update; Railway, Cloudflare, 1Password CLI
Assessment
- Screening call (30 min) → time-boxed take-home (paid for contractors) → live brief-writing exercise (60 min) → founder conversation → references and right-to-work
Product Manager — Agentic AI Systems
Kenjin has a working production framework, a written house constitution and live enterprise clients — and two founders doing all the product management. You'll take that over: turning client problems and market signal into specified, scoped, measurable products, running discovery with clients, and deciding with evidence what ships and what waits.
- Product management of AI/ML or data-intensive products in production
- Can specify agents, memory tiers, RAG, evaluation and human-in-the-loop gates precisely enough to build from
- Enterprise/B2B delivery — discovery workshops, pilots with parallel-run measurement, phased deployment
- Media, entertainment, sports, streaming, archive or talent-industry context — or proof you acquire domain fast
Full role details
What you'll do
- Own the product portfolio across our metadata, enterprise systems, agentic pipeline and audience-intelligence lines
- Run discovery and architecture with clients — workflows, agent roles, memory, integration points, governance boundaries
- Write specs an engineering team or AI coordinator can build from without guessing
- Own metrics: throughput, quality, model performance, human-override rates, time-to-value
- Own the usability bar with the design lead — every surface serves a capable stranger
- Shape commercial packaging and feed the roadmap with market intelligence, separating signal from hype
Bonus
- Multi-source data products; audience or social-media measurement
- Data science literacy — reading an evaluation, a confusion matrix, a retention curve, and challenging them
- Metadata and content standards, C2PA / content authenticity
- Usability and accessibility standards (ISO 9241-11, WCAG); UX audit experience
- Consulting or transformation background in media, sports or entertainment
Assessment
- Screening call (30 min) → take-home spec exercise → live discovery case (60 min) → founder conversation → references and right-to-work
How to apply
No cover letters. Send the thing that actually shows how you work: engineers, the most complex unattended system you've shipped and what it taught you; product managers, a decision you got wrong, what the evidence said, and what you changed. If you meet most of the must-haves and can prove the rest fast, apply. Kenjin is an equal-opportunity employer.
Don't see your role?
If you build or ship the kind of systems we do and think you'd fit, tell us anyway.