Work

Selected work

Products our engineers scoped, built and still support. Client names are withheld where the contract asks for it. What is not withheld is the shape of each build, what it replaced, and which part was actually difficult.

Want to see something closer to your problem? · Reply within 24 hours

SOAP NOTEBAND6.5PAYMENTmatched to billRANKED LEADS
Typical build6 to 12 weeks
Team size1 to 3 engineers
Still supportedAll of them
SectorsHealth, fintech, sales, HR, education
Client namesNamed where agreed
The builds

Different sectors, the same shape of engagement.

One engineer close enough to the business to argue with the brief, a first deploy early enough to be wrong cheaply, and nobody to hand over to afterwards.

CONSULTATIONDoctorPatientNothing typedSOAP NOTESOAPChecked, ready
Healthcare · Clinical AI · Live as Talk!t, Norway

Consultation Summarizer

Turning a live doctor and patient conversation into an accurate transcript and a structured SOAP-format summary, without the clinician typing anything.

Before
Clinicians wrote their notes after the fact, from memory, at the end of a shift. The record was thin, late, and occasionally wrong in ways nobody could audit.
What we built
A speech pipeline that handles overlapping speakers and clinical vocabulary, feeding a language model constrained to the SOAP structure so the output lands in the fields the record system expects rather than as free prose. Handling was built to HIPAA and GDPR requirements from the first commit, not retro-fitted before launch.
The hard part
Clinical language is exactly where general transcription falls over, and a summary that invents a symptom is worse than no summary at all. Most of the engineering went into the checks around the model, not the model itself.
Speech to textLLM pipelineSOAP structuringHIPAAGDPR
Read the full case study →
WRITING TASK 26.5BANDwith reasonsTaskCoherenceLexicalGrammar
Education · Assessment

IELTS Preparation Platform

A learning platform that generates practice material, scores what candidates write and say, and shows them where the marks actually went.

Before
Preparation meant buying a book and guessing. Feedback, when it existed, came from a tutor days later and rarely explained the band score.
What we built
Question generation across the four papers, automatic scoring models for writing and speaking that return a band with the reasoning attached, vocabulary tools that adapt to what the candidate keeps getting wrong, progress tracking, and a full administrative panel for the people running cohorts.
The hard part
Scoring is the whole product. A model that marks generously is pleasant and useless, so the scoring was calibrated against real examiner marks before anything shipped to candidates.
Web appScoring modelsContent generationAdmin panelProgress tracking
Read the full case study →
SpreadsheetEmail approvalsPivot tableONE RECORDSalaryLeave approvedAudit trailSame number
Enterprise · Internal tools

HR Management Portal

One system of record for employees, approvals and reporting, replacing the spreadsheet and email workflow it grew out of.

Before
Employee records lived in a shared spreadsheet, approvals lived in email threads, and reporting meant somebody rebuilding the same pivot table every month.
What we built
Employee records with a proper permission model, approval flows that carry their own audit trail, and reporting built on defined views so that two managers asking the same question get the same number. Delivered as a normal web application their own IT team can host and maintain.
The hard part
The difficulty in internal tools is never the software. It is that the existing process has undocumented exceptions everyone relies on, and finding those means sitting with the people who do the work.
ReactJSNode.jsSQLRole permissionsReporting
Read the full case study →
ACCOUNT · WHY NOWHiring a data lead12 MARNew funding round02 MARGates5 of 5A person decidessources kept
Zeki product · AI sales intelligence

DealBell

DealBell finds companies with two dated, sourced reasons to buy now, and keeps the evidence for every claim.

Before
Reps chased lists with no dated reason to call, and research took an hour per account.
What we built
A sales intelligence engine that ships only accounts with two dated, sourced reasons to buy, with the evidence attached.
The hard part
Making a model that is good at finding reasons refuse to invent them.
Signal intelligenceEvidence trailGatingTypeScriptPostgres
Read the full case study →
Vendor missed datesNeed a mobile teamSwitching vendorRANKED1review + hiring2review + news3watch
Zeki internal · AI lead generation

Competitor Signal Lead Engine

An engine that reads what buyers say about vendors in public, then finds the companies most likely to need help next.

Before
BD relied on bought lists with no intent and no context.
What we built
An engine that reads public reviews and live web signals and ranks companies by a current, sourced reason to buy.
The hard part
Matching the same company across messy sources, and not raising a lead on one weak signal.
LLM extractionSignal intelligenceEntity matchingLead ranking
Read the full case study →
LEADS BY SCORE62544127Call transfervia Twilio
Sales · AI lead scoring

AI Lead Management CRM

A custom lead management system with AI lead scoring, built to replace a legacy platform that had become inflexible and insecure.

Before
A legacy lead system that was inflexible, insecure and worked leads in arrival order.
What we built
A custom CRM with AI lead scoring, AWS serverless endpoints and Twilio call transfer.
The hard part
Migrating years of lead history without losing or corrupting any of it.
PythonDjangoReactJSAWS LambdaTwilioAI scoring
Read the full case study →
BILLdueScan to payMATCHEDPayment inInvoice 1042
Fintech · Payments platform

ZekiPay

A payments platform that lets businesses accept, collect and manage payments in one place.

Before
Recurring collections reconciled by hand, invoice by invoice.
What we built
A payments platform with digital bills, dynamic and static QR, request to pay, an API and a reconciliation portal.
The hard part
Matching every payment to the right bill automatically, across channels.
PaymentsQRRequest to payAPIReconciliation
Read the full case study →
ONELOGINCoursesBookingsStorefrontEmailMembershipsCommunity
Creator economy · SaaS

Creator Commerce Platform

An all-in-one platform for creators, coaches and small business owners to sell and run their business from one login.

Before
Creators stitched together a course host, booking tool, email platform, community app and payment tool.
What we built
One platform for selling courses, products, coaching and memberships, with email, client management and an AI offer writer.
The hard part
Keeping every module simple while sharing one customer record across all of them.
SaaSCourse builderBookingsEmail automationAI copy
Read the full case study →
BDpipelineFollow up todayENGINEERINGPortal v2Mobile appData syncBilling
Zeki internal · Operations

BD and Engineering Dashboards

Two internal dashboards, one for business development and one for engineering, each built around how that department actually works.

Before
Each department ran on spreadsheets, chat threads and status meetings.
What we built
A BD dashboard and an engineering dashboard that put each team's work and conversation in one place.
The hard part
Showing what people already do instead of asking them to report it twice.
DashboardsOperationsTeam communication
Read the full case study →
More projects

Also built by our engineers.

BalanceTransferReceiptPIN2FABiometric
Banking · Mobile · Maldives

Mobile banking app for Maldives Islamic Bank

A customer mobile banking app for Maldives Islamic Bank (MIB), the first Islamic bank in the Maldives. Balances and history, transfers to MIB and other local banks, shareable receipts, cheque services and financing details, protected by PIN, biometric login and two-factor authentication.

POLICYQuote ready
Insurance · Insurtech

Insurtech solutions for SelectQuote

Insurtech engineering for SelectQuote, the US direct-to-consumer insurance distributor. Full details on request.

Clients

Some of the teams we have built for.

Named where the client has agreed. Many contracts ask us not to name the client, so this list is a small part of the work.

SelectQuoteMaldives Islamic Bank (MIB)Talk!t (Norway)Ipsen USVolera Himalayan SaltProspector Metals Corp.Malta Retail Solution
How a build runs

The first deploy lands before the spec would have been signed off.

Every one of these reached a real environment in about a fortnight. Not because the work was small, but because an early deploy is what turns assumptions into corrections while they are still cheap.

The interesting part is the last node. Nobody hands the code to a support desk.
What clients said

Three people who agreed to be named.

“

Each developer I had worked with from Zeki Experts was talented, efficient, communicative, and self-starting.

SWStuart WashingtonProject Manager · Ipsen US
“

Working with the team at Zeki Expert Solutions has been a pleasure. They are professional and efficient.

SSerinaCTO · Volera Himalayan Salt
“

Zeki Experts’ team was exceptional: talented, efficient, communicative, and self-starting.

AJAlicia JackDirector · Prospector Metals Corp.
Questions

The things people ask about this page.

Because most of the contracts say so. Companies that hire engineers to build the thing their business runs on tend not to want that advertised, and we would rather keep the work than the logo. The three named references on this page agreed to be named.

Yes. Ask on the first call and we will arrange it, with the client’s agreement, for work that resembles yours. A reference who built something unrelated to your problem is not much use to either of us.

Yes. The consultation summarizer (Talk!t), the IELTS platform and the HR portal are still running and supported, and every other build shows its current status on its own page: DealBell, for example, is pre-launch. That is the part of the model that matters: the engineer who built it is the engineer you call when it breaks, rather than a support desk reading a handover document.

You do, in your repository, from the first commit. No escrow arrangement, no licence back to us, and no component that only works while you keep paying us.

Next step

Bring us the problem, not the spec.

Thirty minutes, no commitment. If your problem looks like one of these we will say so, and if it is new to us we will tell you how we would approach it.