Where the work I have already done travels
Below is a list of sectors I would happily take a permanent full-time role inside. Each one is mapped to the open-source repositories I have already written that point in that direction. Not a market positioning exercise. A list of places I would be useful if your company is hiring.
Ten sectors I would take a seat in
Permanent full-time PAYE roles only, in the United Kingdom, until at least February 2030. The cards below show where the work I have already done points and what evidence is on the public record.
Fintech, payments, lending
Multi-engine failover, idempotent webhooks, audit logs, and row-level security are the table stakes of any payments stack. I have written the same reliability primitives for SarmaLink-AI because a dropped LLM call is the same shape of failure as a dropped transaction notification.
- SarmaLink-AI: 14-engine failover ladder, circuit breakers, HMAC-SHA256 webhook verification
- Supabase RLS schemas tuned for multi-tenant financial data
- OpenTelemetry traces on every request so a missing transaction is locatable in minutes
Healthtech, med-tech
Audit trails, role-based access, and deterministic replay are non-negotiable. agent-orchestrator was designed exactly around those constraints. Every workflow step is journalled, idempotent, and replayable rather than fire and forget.
- agent-orchestrator: Postgres journal, deterministic replay, BullMQ step queue
- RBAC plus row-level security on Supabase for PHI-adjacent data
- OCR-to-typed-JSON pipelines for intake forms via the receipt-scanner pattern
Retail, e-commerce, ops
I have written the kind of internal portal that retail operations actually depend on. Inventory views, lifecycle tracking, role-based dashboards in one Next.js app on Supabase. I know what breaks at the catalogue and fulfilment layer under real load.
- staff-portal: internal ops template with lifecycle tracking and role-based dashboards
- webhook-to-email for order events, low-stock alerts, failed-payment notifications
- AI assistant with product-catalogue retrieval over private documents
B2B SaaS and developer tools
This is the home sector. Nineteen open-source repositories spanning coding-agent runners, multi-provider gateways, Rust inference servers, multi-tenant SaaS scaffolds, eval runners, MCP servers, voice agents, LSM-tree engines, Raft KV stores, and WebAssembly sandboxes. Multi-tenant architecture and developer experience are the defaults I think in.
- shipyard: tenant isolation, RBAC, billing, audit log, rate limits in one TypeScript stack
- ai-eval-runner: evals as code with DuckDB result store and HTMX viewer
- mcp-server-toolkit: production MCP server starter in Python and FastAPI
Edtech and e-learning
Real-time voice and retrieval over private documents are the two primitives that power personalised tutoring. I have shipped both. voice-agent-starter holds a sub-second round-trip; rag-over-pdf handles chunked ingestion with citation spans that survive into the answer.
- voice-agent-starter: WebRTC plus mediasoup, sub-second round-trip
- rag-over-pdf: citation-first retrieval over course material
- Multi-engine failover so a model outage does not take the learning tool offline
Media, publishing, content platforms
This site itself is the reference. Long-form posts with charts, comparison tables, citations, and live-data widgets rendered from a Supabase-backed content pipeline I wrote and maintain. The same pattern would carry into an editorial CMS for a real media team.
- Rich content renderer with charts, comparison tables, citations, and live widgets
- Live API widgets pulling from public APIs with a 30 to 60 minute cache
- AI-assisted drafting with multi-engine failover so the writing tool stays up
Logistics and supply chain
Intent classification, structured-JSON responses, OCR over shipping documents, and a platform that scales with shipment volume spikes. The pattern is the same as for any track-and-trace surface: classify the request, dispatch to the right source, return a typed response.
- Plugin auto-routing inside SarmaLink-AI with intent classification
- terraform-stack for Vercel plus Supabase plus Cloudflare plus DigitalOcean modules
- receipt-scanner architecture for customs documents and shipping paperwork
Govtech and the public sector
Open-source by default, MIT licensed, runbooks living in the repository next to the code. Auditable workflows on Postgres with deterministic replay and structured logs. The shape a public-sector platform actually deserves.
- agent-orchestrator: durable workflows with full replay for accountability
- k8s-ops-toolkit: ingress-nginx, cert-manager, kube-prometheus-stack, Loki on day one
- Eighteen MIT-licensed repositories so nothing I bring is paywalled
Telecoms, real-time, infrastructure
Sub-second voice loops, paged KV-cache inference, LSM-tree storage engines, and a Raft KV store with a fault-injection harness proving linearizability under partitions. The low-level layer the average product team never gets to write.
- voice-agent-starter: full-duplex WebRTC loop with per-stage latency telemetry
- forge-infer: paged KV cache, continuous batching, speculative decoding in Rust
- lsmdb plus raftkv: the storage and consensus layer beneath any serious system
Charities and non-profits
I do pro-bono engineering for charitable organisations on the same stack I use commercially. Next.js, Supabase, Vercel runs at near-zero marginal cost for low-traffic non-profits, and what gets handed over is a maintainable codebase, not a black box.
- Full Next.js plus Supabase plus Vercel stack on minimal hosting cost
- Staff portal and CRM-lite for volunteer management and donor tracking
- Automated comms via Resend for receipts, event confirmations, newsletters
Sector is rarely the blocker. Permanent PAYE is.
I am open to permanent full-time PAYE employment in the United Kingdom only, and only until at least February 2030. Employee roles only. If a sector above maps to your team, the hire-me page has the role shapes, the capability matrix, and the email that reaches me directly.
Read the hire-me page