Software & Applied AI Engineering
I build the tools operators use to find the right policy, review a request and recover a failed delivery.
I take it from the first operator workflow
through release and handover.
One enquiry.
One CRM record.
Replay a webhook, interrupt delivery, then retry it. The operator can see the failure and recover the approved enquiry without creating a second record.
Self-initiated demo · fictional data · CRM simulator
Select “Run recovery demo”. After the retry, compare the event history with the one exported record for that enquiry.
AI platform
Search that traces a passage back to its source.
I work on ingestion, retrieval and AI-provider integration. In an offline check, updating a sample returns policy from 14 to 30 days changed the retrieved passage. A question about opened parcels still retrieved a rule for unopened parcels. That is a case I would test before putting a support assistant in front of customers.
Data pipelines / retrieval / RAG / provider integration
Platform project · public demo unavailable
Mathematics · AI interface
A mathematical notebook built around handwriting.
Handwritten input and AI features in a single product. My role covers requirements, prioritisation, user scenarios and development coordination.
Product requirements / user journeys / delivery coordination
Product project · overview
Android · cloud storage
File transfer and backup to S3-compatible storage.
An Android application for moving and backing up files. My role covers requirements, priorities, user scenarios and development coordination.
Android product / S3-compatible storage / user workflows
Product project · overview
Engineering scope
Product thinking and operational experience, backed by hands-on engineering.
Turn an ambiguous business problem into requirements, architecture and a working product. Own the decisions across backend, interface, integrations and deployment.
Build retrieval into the product and test it on questions people actually ask. Trace each cited passage to its source and check when the assistant should abstain.
Structured logs, traces and relevant metrics. Make failures diagnosable, recovery deliberate and quality, latency and cost measurable where they matter.
Java · Python · Go · JavaScript · Vue
The engineer
Software & Applied AI Engineer
with product and operational experience.
My background connects enterprise software, insurance workflows, product development and operational work with clients and contractors. I work across the business process and the system that supports it.
Medical assistance software for operators and insured users: organising care, arranging appointments and connecting people with services. Organisation names and internal details remain confidential.
I use AI agents to accelerate development. Technical decisions, quality checks and commitments to the client remain my responsibility.
A focused way to start
Send a webhook that creates two records, a question your search gets wrong or a task your team still does by hand.
We use one example to agree the expected result, the cases to test, the fee and the delivery date.
I implement the change, rerun the example and show where your team can see stalled work or unsupported answers.
You get the code, test results and a short guide to running the feature and handling the failures we covered.
Quoted after a written brief. If the system needs investigation first, we agree a separate discovery stage. Additional workflows and ongoing support are scoped separately.
See the three ways to work togetherLet’s put it to work
Send what exists, what should happen and the constraints.
I’ll reply with scope questions and a concrete first step.
No call required. Please keep credentials and customer personal data out of the initial brief.
