Each area is a terminal full of concrete deployments — from customer service and sales agents, through self-hosted LLM and RAG, to data pipelines, lead scoring, and offer portals. We start with an audit and pilot: pick one process, demonstrate a working system, and only then discuss scaling. Access to models always goes through a router, with PII masking before the cloud.
// 00Service catalogue
What you can build with AI — 110+ services.
Over 110 AI services across 7 consoles — from agents and infrastructure, through data, voice and security, to apps, PropTech and deployments.
++++
07consoles
110+services
// 01Service consoles
cashcrown@lab: ai.agents --listready
Autonomous agents & automation
01/07Agents that do real work inside real processes — support, sales, documents, integrations — and report the truth, not optimism.
- $collections / payment-reminder agent - polite reminders, escalation on your thresholds, hard cases to a human
- $meeting-scheduling agent - finds calendar openings, proposes slots, confirms — conflicts go to a human
- $content-moderation agent - flags comments and submissions by your rules; borderline calls go to a moderator
- $candidate pre-screening agent - sorts CVs and intake answers against your criteria; the recruiter decides who to invite
- $monitoring & alerting agent - watches metrics, price changes or mentions and pings only when it actually matters
- $multi-agent orchestration - a coordinator agent splits work across specialist agents and keeps the result coherent
- $contract-review agent - flags risky clauses and deviations from your template; the final call stays with a lawyer
- $invoice-processing agent - reads invoices, matches them to orders and pre-posts them; mismatches go to human approval
22 services in „Autonomous agents & automation”whole category
cashcrown@lab: ai.infra --listready
Sovereign AI infrastructure
02/07Your models, on your hardware. Self-hosted LLMs, RAG, gateway and observability — with predictable cost and no lock-in.
- $on-prem GPU server - we set up and harden the inference box
- $model containerization - Docker / Kubernetes, reproducible deploys
- $GPU monitoring & alerts - load, temperature, failures — before they hurt
- $inference autoscaling - capacity scales with traffic, not the bill
- $AI backup & disaster recovery - models, vectors and config restorable — with a restore test
- $on-prem STT / TTS - speech runs locally, recordings stay in-house (GDPR)
- $OpenAI-compatible local API - same endpoint, your code never notices the swap
- $maintenance & SLA - updates, on-call and an agreed response time
20 services in „Sovereign AI infrastructure”whole category
cashcrown@lab: ai.data --listready
Data, knowledge & visibility
03/07From data acquisition and cleaning to embeddings, SEO and AEO. Visibility in Google and in model answers.
- $data cleaning - standardization, gaps, typos — with a report of what changed and why
- $deduplication & record matching - merges the same companies/people across sources, thresholds for human review
- $document digitization (OCR) - scans and PDFs → structured data, accuracy in ranges, human verification
- $audio & video transcription - recordings → timestamped text, run locally when data is sensitive
- $data labeling for ML - training annotations with quality control and double-checking
- $knowledge graph - entities and relations from your data, a base for RAG and search
- $anonymization & pseudonymization - anonymization takes data out of GDPR scope; pseudonymization reduces risk but data stays in scope — we match the method to the use case
- $data quality monitoring - rules and alerts for anomalies before they reach a report or model
22 services in „Data, knowledge & visibility”whole category
cashcrown@lab: ai.build --listready
Apps, PropTech & integrations
04/07Property-tech and apps built like a product: listing portals, CRM sync, auto-generated pages, APIs and visualizations.
- $listing feed syndication - one feed → Otodom, OLX, Gratka, your own site
- $lead routing to agents - assigns each enquiry by duty roster and specialty, logged
- $viewing scheduler - proposes slots from your calendar, confirms, reminds
- $contract & document intelligence - extracts key fields from contracts and deeds; the call stays human
- $listing photo cleanup & virtual staging - straightening, exposure, furnishing empty rooms — reviewed before publishing
- $market & comparative reports (CMA) - price ranges from comparable listings, with data sources
- $tenant & buyer pre-screening - ranks applications by transparent criteria, GDPR-compliant; final call is the agent's
- $owner valuation lead magnet - a "what's my home worth" calculator → a qualified lead, with consent
22 services in „Apps, PropTech & integrations”whole category
cashcrown@lab: ai.voice --listready
Voice & telephony AI
05/07Voicebots, transcription and call assistance — honest about Polish ASR limits, with a handoff to a human on hard calls.
- $inbound voicebot - answers, qualifies, escalates to a human
- $outbound voicebot - reminders and confirmations, with consent
- $IVR replacement - a conversation instead of “press 1” menus
- $appointment-setting callbot - proposes slots and books them into the calendar
- $call transcription - transcript and summary of the call into the CRM
- $voice analytics - topics, intents and service quality
- $real-time agent assist - prompts and answers from the knowledge base mid-call
- $after-call work automation - notes, tags and tasks without manual clicking
10 services in „Voice & telephony AI”whole category
cashcrown@lab: ai.secure --listready
AI security & compliance
06/07Red-teaming, guardrails and readiness for GDPR and the AI Act — with a human in the loop. We reduce risk and document it honestly; we don't promise 100% safety.
- $LLM red-teaming - we attack the model before someone else does
- $prompt-injection audit - we test injections and instruction leaks
- $GDPR & AI Act readiness audit - gaps and an action list before you ship
- $DPIA support - data protection impact assessment, with your counsel
- $guardrails layer - input/output filters and hard rules
- $PII masking & pseudonymisation - only a masked prompt leaves for the cloud — no raw personal data
- $AI assistant security audit - we review the bot, its tools and permissions
- $abuse & anomaly monitoring - alerts on attacks, cost spikes and odd traffic
10 services in „AI security & compliance”whole category
cashcrown@lab: ai.ops --listready
Deployment & MLOps
07/07The road from pilot to production without surprises: versioning, golden-set quality gates, cost and latency monitoring. Every deploy passes an eval gate and a human sign-off.
- $pilot → production - turning a demo into a stable deployment
- $LLMOps / MLOps from scratch - CI/CD, model registry, pipeline
- $prompt & model versioning - you know what runs and what changed
- $golden-set eval gates - a regression blocks the deploy, not the client
- $quality, cost & latency monitoring - alerts before quality drops
- $on-prem / self-host rollout - on your hardware, your data stays with you
- $canary / shadow rollout - a new version on a fraction of traffic
- $incident runbooks & response - what to do when the model fails
10 services in „Deployment & MLOps”whole category
// 02Next step
++++
Don't see your process? We'll build it.
The catalogue is a starting point, not a ceiling. We begin with an audit and a pilot.
Book a call