Solutions
Different industries, the same bottleneck
We only take on cases where the method is already proven. These three are our current focus; screening and pharma have their own page.
New energy & process manufacturing
Capture the veteran operator's judgement in software: setpoint decisions at the optimisation layer, process-to-property models, electron-microscopy quantification, industry intelligence agents, and a web system managers can actually read.
Read more →Higher education
Rules spread across a dozen pages, converged into one answer that can be cited and audited; one platform carrying many faculty instances, with student data that never leaves the country.
Read more →Hospitals & departments
MedLoop: the clinician sets the plan, AI runs it day to day, and only what needs clinical judgement comes back to the doctor.
Read more →Screening & pharma
Who catches the patient in the three years after a positive result — screening, diagnosis, adherence and real-world evidence as one measurable loop.
Read more →02 / Higher education
One handbook across a dozen pages, and still only one right answer
Rules are scattered, answers differ by window and by day, peak season overwhelms staff — and a subset of questions simply cannot be wrong: deadlines, fees, appeals, graduation requirements, academic standing. Getting those wrong carries an institutional cost.
Working well is not the same as being deployable. Any outsourced solution also has to clear data governance, IT and procurement review. We treat that as a design constraint from the first prototype page.
Our red line: we never promise 100% accuracy, and we never sign a numeric threshold before seeing real traffic.
Change the school and you change the content, not the platform
- The change-control and rules-layer modelling method — one handbook, many programmes, each with exceptions
- A cited, refusable, human-in-the-loop answering framework
- Data and acceptance language that survives governance review
- The same base moves to admissions, registry, international student affairs, graduate school, housing and financial aid
Staged release, widened step by step
Stage 0: staff only — the system drafts, a human sends. Stage 1: limited student release, and only categories that passed acceptance may auto-send. Stage 2: whether to widen is the faculty's call, made on real data.
High-consequence questions always go to a human
Deadlines, fees, appeals, graduation requirements, academic standing, rule conflicts, and any sign of urgency or distress — 100% routed to staff. The client may extend that list but never shorten it.
Not "AI confidence" — a calibrated routing score
Computed from retrieval coverage, citation support, rule conflict, out-of-scope detection and question category, and calibrated on an evaluation set. It is not the model judging itself — that sentence moved the technical review from sceptical to supportive.
Make cross-border transfer impossible by design
Student personal data stays in the Singapore region; three separate cloud accounts with region locked by organisation policy; the offshore team only ever touches synthetic and de-identified data. We distinguish de-identification, pseudonymisation and anonymisation strictly — and we do not claim anonymisation.
Acceptance you can adjudicate, and what happens if we miss
Material error is judged by the client's content owner and that judgement is final; auto-send accuracy is sampled weekly; a repeat contact on the same issue within seven days counts as unresolved. If we miss, the client picks one of four: stay in draft mode, narrow the auto-send categories, let us remediate free within two weeks, or stop and pay only for what was accepted.
03 / Hospitals & departments
Clinicians don't need more data; they need an executable loop outside the hospital
Once the consultation ends, no clinician can keep every patient on their medication, appointments and daily behaviour. As panels grow, follow-up calls, screening and report questions eat the day. The gap between two visits goes unrecorded — and raw data without risk stratification only adds load.
Open the MedLoop pageContact
Start with one conversation about the case
If you have a specific case — one production line, one faculty, one department, one clinical pathway — we'll do a feasibility conversation at no charge and tell you whether it's worth building.