A local AI that parses entire Acts and improves with every use
Big Four-level advisory, without the Big Four price
Index visualization only; not tax advice
ACT BRAININTERACTIVE INDEX
/
Search a section to travel its pathDrag to pan
Document intelligence
The document becomes the data.
Watch a statement become verified, structured accounting entries.
Document parsingReconciliationAutomation
Receive the statement
Align the page
Detect the table
Identify columns
Locate words and amounts
Check the running balance
Build structured rows
Prepare accounting entries
DOCUMENT PARSER
Healthcare
A schedule is a plan. The room tells the story.
Built on my perioperative analytics dashboards: booked time compared with what actually happens.
DE-IDENTIFIED PDF PATTERN No patient identifiers shown.
Operating room performance / two rooms
Same cases. 30-minute turnovers.
Two rooms from Day 1. Every case keeps its duration; only the turnovers change, to 30 minutes.
Rooms in use--
Longest turnover--
Longest free block--
Time runs down from 07:00 to 19:00; shading marks time after the 07:00-17:00 staffed block. Moving cases between rooms needs compatible staffing, equipment and patient readiness.
Operating room performance / all 17 rooms
Scaling the problem.
The same rule across all 17 rooms, each day as it happened and then rearranged.
Day 1
Room utilization--Actual time / 07:00-17:00 capacity
Duration accuracy--Duration within 20 minutes
Time after 17:00--Total room minutes beyond block
Completed cases--Across the same 17 rooms
ScheduledWithin 20mShorterLonger
Outlined = plan. Colored = actual. Select a case to compare.07:00-17:00 staffed block
Select any colored case for its scheduled and actual times.
40 bills from one procedure class; each of the six fixed and three variable items appears in 36 (90%). Red marks recurring variable charges, not errors or fraud. This item-frequency view is separate from the 500-case distribution below, and no patient records or identifiers are used.
Estimate before the bill arrives
From 500 cases to a clearer range.
Fixed charges versus varying equipment and disposables. Each dot is one bill from the same procedure class.
Colored dots show the current step; grey dots the other bills.
One case90% next-case range95% confidence interval for mean
Average of analysed cases
95% confidence interval for average
90% estimated range for a similar next case
Who covers what
The same range, seen by five payers.
Each row is an insurer or payment mode. Band: where its bills fall. Bar: what cover pays before the cap. Dot: average company share; tick: average bill.
Payer names withheld; no insurer, claim or patient identifiers are shown.
PatientPlan around a range and see which items move the bill.
HospitalSeparate the fixed base from equipment and consumable variation.
InsurerAssess bill variability before applying coverage terms and patient share.
How the ranges are calculated
500 cases, one procedure class: fixed charges are INR 75,000 per case; equipment and disposables vary.
The narrow band is a 95% percentile bootstrap confidence interval for the average (400 resamples), not for one patient's bill. The wider range is a 90% split-conformal prediction interval (80% baseline, 20% calibration) that assumes future cases follow the same exchangeable process; it has not been validated on real hospital bills.
More cases sharpen the average, not the variation between patients. Shading shows the central share of observed amounts, not a guarantee; equipment and consumable amounts are bill components, not verified costs or margins.
No original patient, doctor, hospital, insurer or claim identifiers are reproduced.
From static to interactive
The original work connected utilization, case-length accuracy and room timelines; this view rebuilds 17 room lanes from a dashboard PDF, with approximate timings and all names, dates and case identifiers removed. Capacity views are not clinical recommendations.
Every case keeps its duration and each room keeps a 30-minute turnover. Live scheduling would also respect service-reserved rooms, surgeon and anaesthesia team availability, sterile instrument turnaround and emergency time.
Manufacturing / Lean Six Sigma / AI & automation
Connect the floor. Improve the flow.
Industrial engineering makes the whole system work better. My Lean Six Sigma Green Belt training pairs process mapping and root-cause analysis with software that makes improvement repeatable.
Process improvementHardware integrationEmployee self-service
Biometric device → attendance → payroll
From a punch to a clear record.
A workflow I built. It connects existing hardware to software and lets people resolve exceptions the same day.
Attendance registerTuesday
Employee
Check in
Check out
Hours
Status
Our software / the vendor's hardware
The vendor's hardware. Our own rules.
We connected our own software directly to the vendor's biometric hardware. Data, rules and reports now follow our process, with no hand edits.
I can build the same for your organization.
Data collectionRead from the device, kept untouched.
ComputationHours and status calculated by software.
ReportsPayroll sheet prepared automatically, then reviewed.
ProcessExceptions go to the employee as marked corrections.
Small details. Real consequences.
Preserve the original.
Corrections are stored apart from device punches, so changes stay visible.
Resolve it while it is fresh.
Give employees visibility.
Staff review attendance and record missing punches during the month, not at payroll time.
Improve before you automate.
Reduce the repeat work.
Map handoffs and validate inputs, then automate the stable steps.
Where this approach can go next
AI and automation for the manufacturing workflow.
I can connect machine events to dashboards, extract production records from documents and route exceptions to the right person.