Skip to content
Built by SamirTalk to Samir →

Proof of work

The Lab

Consulting advice ages badly when the person giving it hasn't shipped anything recently. So I keep building — production systems, real infrastructure, on my own time and my own money.

These aren't products I'm selling you. They're how I stay current, and how you avoid paying for my learning curve.

Five are documented in depth below, one project per page: the real stack, the real architecture, and one genuine engineering decision from each — including the ones that show a project stalling, a model getting shut off, or a feature shipping dark because it wasn't ready. That's the point. If I only showed you the parts that worked, this page would be a sales pitch instead of evidence.

Where I build and test before clients pay for it.

Capability matrix

What maps to what

Six capabilities, and the operational symptom that makes each one relevant to you.

Capability matrixA six-row matrix mapping each capability — voice AI and phone handling, predictive modeling with honest validation, computer vision and extraction, third-party API and marketplace integration, autonomous agent orchestration, and consumer mobile delivery — to the client situation it is relevant to.CapabilityRelevant whenVoice AI andautomated phonehandlingfront desk overload, missed calls,after-hours coverage.Predictive modelingwith honest validationsomeone is about to sell you anAI accuracy number.Computer vision andAI extractiondocument intake, chartabstraction, manual data entry.Third-party API andmarketplace integrationyour systems don't talk, or anintegration could do somethingyou can't undo.Autonomous agentorchestrationrepetitive knowledge workat volume.Consumer mobile andend-to-end deliveryyou need someone who finishes,not someone who prototypes.

Voice AI and automated phone handling

Built a complete AI receptionist platform: inbound call handling, appointment booking, call transcription and action extraction, telephony and payment integration.

Relevant when: front desk overload, missed calls, after-hours coverage.

Five projects, one page each

The Lab, project by project

Each write-up carries the stack, the architecture, and the one decision I'd defend in a review.

TruLine's validation chart, compact viewA small line chart with no numeric values: a line rises confidently, then at a marked point a corrected line drops below the starting baseline, showing an honest re-score inverting a favorable-looking result.
01 / 05

TruLine

prediction modeling and honest validation

Explore the build
02 / 05

CardScope

computer vision extraction and a third-party integration with an irreversible-action gate

Explore the build
03 / 05

Cadence

consumer mobile, built and delivered end to end

Explore the build
TicketScan's agent fleet, compact viewA central scheduler node connected by eight thin lines to eight small agent nodes arranged in a ring, illustrating a fleet of scheduled agents run from one orchestrator.CRON
04 / 05

TicketScan

a fleet of autonomous agents that run unattended

Explore the build
Stock Screener's decision pipeline, compact viewA horizontal pipeline of three stages — screen, classify, rule gate — where the rule gate branches into two outcomes: execute, continuing forward, and block, terminating the path. No figures or performance numbers are shown.SCREENCLASSIFYRULE GATEEXECUTEBLOCK
05 / 05

Stock Screener

rules-driven decision automation

Explore the build