You will get Automated Audit Pipeline with AI Risk Reports | Procurement & Beyond
Project details
This isn't a generic "AI consultant" pitch. I'm the Head of Purchasing at a large pathology lab network — I ran the procurement function before I automated it. The system you're looking at audits 30,858 POs in production, every morning, across 3+ years of history. 22 rules catch rate anomalies, indent mismatches, workflow violations, and receipt exceptions. Each PO gets a weighted risk score; your audit team wakes up to a ranked-priority queue of ~300 HIGH-risk POs—not 189,000 raw flags.
What makes this different: (1) I built it for my own team first, so the rules reflect real fraud and leakage patterns, not textbook ones. (2) No cloud lock-in — runs on a single machine, extracts overnight, and the infrastructure cost is effectively zero. (3) You get the full stack: ETL pipeline, 22-rule engine, risk scoring, Excel workbook, Slack summary card, and a Claude-drafted executive narrative delivered to your inbox. (4) Four-to-six-week delivery, with a two-week pilot run alongside your existing process before handoff.
I bring a background as a Chief AI Officer, plus an Executive MBA from LUMS — I speak both CFO and tech.
What makes this different: (1) I built it for my own team first, so the rules reflect real fraud and leakage patterns, not textbook ones. (2) No cloud lock-in — runs on a single machine, extracts overnight, and the infrastructure cost is effectively zero. (3) You get the full stack: ETL pipeline, 22-rule engine, risk scoring, Excel workbook, Slack summary card, and a Claude-drafted executive narrative delivered to your inbox. (4) Four-to-six-week delivery, with a two-week pilot run alongside your existing process before handoff.
I bring a background as a Chief AI Officer, plus an Executive MBA from LUMS — I speak both CFO and tech.
AI Algorithms
Large Language Model, Transformer ModelAI Applications
AI Content Creation, Anomaly Detection, Natural Language Generation, Time Series AnalysisAI Development Language
PythonAI Models
GPT-4What's included
| Service Tiers |
Starter
$1,500
|
Standard
$3,500
|
Advanced
$5,500
|
|---|---|---|---|
| Delivery Time | 14 days | 28 days | 45 days |
Number of Revisions | 2 | 4 | 10 |
AI Model Integration | |||
Batch Normalization | - | - | - |
Database Integration | |||
Detailed Code Comments | - | ||
Image Upscaling | - | - | - |
MLOps | - | - | |
Model Deployment | - | ||
Model Documentation | |||
Model Monitoring | - | ||
Model Testing & Optimization | - | - | - |
Model Tuning | - | - | - |
Natural Language Processing | |||
NLP Tokenization | - | - | - |
Pre-Training | - | - | - |
Prompt Engineering | |||
Setup File | |||
Source Code | - |
Optional add-ons
You can add these on the next page.
Fast Delivery
+$400 - $1,800
Additional Revision
+$150
Additional Audit Rule
(+ 3 Days)
+$300Frequently asked questions
About Muhammad
Healthcare AI Consultant | Diagnostics, Clinical Data & Automation
Lahore, Pakistan - 9:03 pm local time
I'm Chief AI Officer at one of the largest diagnostic networks in my region, where I deploy AI and automation across diagnostics, clinical data, and operations — for an organization that serves millions of patients in a tightly regulated environment. So I build for the realities healthcare clients actually face: privacy, auditability, and trust come first.
A healthcare system I built and shipped:
🔹 AI "Digital Patient Twin" — Pulls years of a patient's lab results, scans, and biopsy reports, blends in their Apple Watch / Apple Health data, and turns it into one short, doctor-ready summary: who is this patient, what's changing over time, and what to check next. It runs several AI models in parallel and has a stronger model write the consensus, with full cost tracking (~15–30¢ per report). Privacy-first by design — the AI never sees a name, ID, or phone, only an internal record number, age, and sex; the database connection is strictly read-only; every report is clearly labelled as clinical decision support, not a diagnosis.
Other systems I've delivered:
🔹 Automated Procurement Audit — Reads an Oracle ERP on a set schedule and scores purchase orders against ~22 audit rules (duplicate approvals, vendor price creep, goods never received, split POs). It runs silently and only surfaces what needs attention — a short, plain-English brief in management's inbox. Read-only end to end, so it runs against live production safely.
🔹 Sales Intelligence Dashboards — Turned a year of messy WhatsApp chats and CRM data into clean HTML dashboards that pinpointed exactly where deals were dying. Transcribed 850+ voice notes locally for privacy.
Through my consultancy, catalase.ai, I bring this operator experience to healthcare, services, and trading businesses moving from "we should probably use AI" to "AI is now part of how we operate."
What I bring:
🔹 Full stack, end to end — data pipelines, AI integration, dashboards, and the scheduling that runs it unattended.
🔹 Safe by default — read-only data access, anonymization, clear AI guardrails. Built for regulated environments.
🔹 Tied to outcomes — time saved, cost reduced, better decisions.
🔹 20+ years across data, operations, and technology leadership, plus an Executive MBA in progress at LUMS.
How I work: transparent scoping, realistic timelines, and working systems over slide decks. If your project isn't a fit, I'll tell you and often point you toward someone who is.
Message me with a rough description of what you're trying to solve, and I'll reply within 24 hours with an honest assessment and, if it's a fit, a clear next step.
Steps for completing your project
After purchasing the project, send requirements so Muhammad can start the project.
Delivery time starts when Muhammad receives requirements from you.
Muhammad works on your project following the steps below.
Revisions may occur after the delivery date.
Discovery & data audit
Stakeholder interviews, ERP schema review, and rule scoping. I map your procurement controls to the 22-rule audit library and confirm which flags matter most for your team.
ERP connection & extraction pipeline
Set up read-only access and build the nightly extraction pipeline from your ERP into a local Postgres warehouse. Incremental loads, idempotent, zero changes to production data.