"We'd put serious engineering into our voice agent and kept hitting a quality ceiling we couldn't break through. Piotr showed us the scope was just too broad for what LLMs can reliably do today. We decided to reframe the product as a focused 24/7 intake assistant, and we ended up with higher customer satisfaction at a fraction of the complexity."LinkedIn open_in_new
LiveKit · Pipecat · Your cloud
Reliable Voice AI you own.
Forward-deployed
voice AI team
for regulated call centers
My team and I work inside yours and build voice agents on LiveKit or Pipecat in your own cloud, so you stop paying per-minute platform fees. If you would rather stay on Vapi or Retell, we improve that deployment instead. I've run agents handling 50M+ calls a year in debt collection, and I cut one client's voice AI pipeline operating cost by 90%.
Get a free consultation arrow_forward50M+
Calls / year in production
30-50%
OPEX cut
GDPR · HIPAA
Built for compliance review
Where are you with voice AI?
We want voice AI, but don't know if it will work.
We start with a pilot on one call type. I build it on LiveKit or Pipecat, test it with simulated calls, and put it on a real phone number with a share of your live traffic. Compliance is built in from the first call. You get real numbers on cost and CSAT before you commit to a full rollout.
Includes A working agent on a real number, the code in your repo, a simulated-call test suite, every call scored, and a cost model with a rollout plan.
Find out more
Our voice AI costs too much or keeps failing.
I build dashboards that show what every call costs and where it fails. Then I fix it with TTS and LLM caching, leaner call flows, models fine-tuned for your calls, and, where your volume justifies it, a move off per-minute platforms to LiveKit or Pipecat in your cloud. For one debt-collection client this cut the pipeline's operating cost by 90%, saving $300k+ a year.
Includes Per-call cost and quality dashboards, a caching layer, model fine-tuning, and latency and hallucination fixes.
Find out more
Legal or security won't sign off.
I map your data flows and build consent capture, call recording, PII redaction and data residency into the system, so your legal team reviews a finished design. One client had four weeks of GDPR work ahead of them and went live in a week.
Includes GDPR and HIPAA architecture, data-flow mapping, and self-hosting in your cloud or region.
Find out more
We're scaling and can't hire fast enough.
I join your team with my own developers and QA testers. We build, test and harden the system in your cloud, set up monitoring, and hand it over to your engineers when they're ready. At a debt-collection operation handling 50M+ calls a year, this cut call-handling OPEX by 30-50% compared with human agents while CSAT held.
Includes Developers and QA on demand, observability, vendor management and handover to your team.
Find out more
Ways to work together.
Start small and grow. Each one is scoped and priced in a working session.
Pilot
Teams that want proof on one call type before they commit.
Scoped in a working session
- What we build
- One call type on a real number with a share of live traffic, a simulated-call test suite, and every call scored against success criteria agreed up front.
- Who's on it
- Me, from scoping to the go or no-go call.
- Where it runs
- Your cloud, on your own LLM, speech and telephony accounts, so you pay providers at cost.
Up to 12 weeks. The pilot fee comes off the production build.
Production build
Call centers moving one line of business to voice AI.
Scoped in a working session
- What we build
- Everything in the pilot, plus more call types and languages, TTS and LLM caching, and models fine-tuned for your calls.
- Who's on it
- Me, with my developers and QA testers when the work needs them.
- Where it runs
- Your cloud, or the platform you already use. GDPR and HIPAA controls where you need them.
The code goes into your repo and is yours once the build is paid for.
Forward-deployed team
Operations at or heading toward 1M+ calls a year, across call types, departments or brands.
Scoped in a working session
- What we build
- Everything in the production build, plus monitoring and alerts, vendor management, and a handover with runbooks.
- Who's on it
- Me and my developers and QA testers, working inside your team until your engineers take over.
- Where it runs
- Your cloud. After handover, your engineers run it, or I keep it running on a monthly support plan.
Team size scales up and down with the work.
Who I work with.
account_balance
Debt collection agencies
Outbound calls at scale with compliant scripts and a lower cost per contact. I've run voice AI handling 50M+ calls a year in this space.
account_balance_wallet
Banks and fintech
Inbound support agents built to the standards I worked under at Kraken, State Street and Argent.
health_and_safety
Insurance
Claims intake, broker support and underwriting Q&A, automated without new audit exposure.
medical_services
Healthcare and telehealth
Scheduling, triage, intake and follow-ups, with HIPAA controls built in from day one.
gavel
Legal firms
Intake, conflict checks and scheduling, with privilege and confidentiality protected.
concierge
Hospitality
Bookings, common questions and pre-arrival calls handled 24/7, so staff can focus on guests.
What clients say about my services.
"Four weeks of GDPR work and deployment decisions in front of us. Piotr walked us through the options, cleared the legal path, and we went live in a week."LinkedIn open_in_new
"Our agent wasn't hitting the quality bar we needed. Piotr from Calloptima walked us through the architecture options with clear trade-offs, shaped a CSAT strategy that actually moved the numbers, and left us with a testing and iteration playbook we could run ourselves."LinkedIn open_in_new
lock
Global client · Debt collection
Under NDA
50M+ calls/year · 30-50% OPEX cut vs humans · 90% pipeline cost cut
"Embedded voice AI lead for a global debt-collection operation handling 50M+ calls a year. I cut call-handling OPEX by 30-50% compared with human agents while maintaining CSAT, then cut the existing voice AI pipeline's operating cost by 90%, saving $300k+ a year."Read the breakdown open_in_new
I'm Piotr Żaczek. I also record video breakdowns of voice AI architectures and what they cost in production.
Why Calloptima?
Most voice AI projects look good in a vendor demo and break in production on latency, compliance or cost.
I built Calloptima to fix that. I've run voice AI operations handling 50M+ calls a year in highly regulated sectors: debt collection, institutional finance and fintech.
I work inside your team, build in your cloud or on the platform you already use, and stay until it holds up at scale. When the work needs more hands, I bring my own developers and QA testers.
FAQ.
Do you use Pipecat/Vapi/Retell/ElevenLabs/LiveKit, etc.? Which one do you recommend?
I pick the framework to fit the client. Usually that is LiveKit or Pipecat, both open source and proven in production. I have run the hosted platforms too, and there are cases where they win on time to market. In the free consultation I go through your requirements and pick the best fit. I also made a video on how I choose based on business needs: watch on YouTube.
Do we have to move off Vapi or Retell?
No. Moving to LiveKit or Pipecat in your own cloud removes the per-minute platform fee, which pays off once your call volume is high enough. If you would rather stay where you are, I work inside your current deployment and fix cost and quality there.
We don't have engineers to run it. Can you?
Yes. On a monthly support plan I keep it running in your cloud: monitoring, fixes, and model or provider updates. The infrastructure stays yours, so if the plan ends, nothing stops working.
Do you build full pipelines? Could you help me with HIPAA/GDPR?
Yes to both: I build the pipelines, and I cover the technical side of compliance. I have built platforms handling hundreds of thousands of HIPAA-regulated calls a month, and I can get you live quickly while keeping conversational quality, a full audit trail of every conversation, and alerts that flag failures before your customers notice them. GDPR runs on the same architecture with its own controls; see GDPR and HIPAA compliance.
How much do your services cost?
It depends on the service, the amount and complexity of the work, and the expected ROI. I have saved $300k+ for my clients. Before any work starts, we agree on a plan with the cost and deliverables in it.
Will it work in my case? I have custom needs I need to accommodate.
Every client I've worked with had a different problem, and all of them were happy with the result. In the free consultation we'll work out how to handle yours.
Who do you help? What do you do?
My team and I work forward-deployed inside mid-size and large companies, building and optimizing voice AI on their own infrastructure. That covers new agents, cost and quality fixes for existing ones, the technical side of compliance (HIPAA, GDPR), and developers and QA testers when your team needs more hands.
What about data privacy and security?
Your agents run in your own cloud, so call data stays under your security policies and your own HIPAA or GDPR setup, not a third-party voice platform's.
Get a free consultation.
A working session on your call flows, your telephony and your constraints. I read every message personally and reply within 1-2 business days.