Skip to content

Delivered projects

  • IT-Megastar

    HR AI agent

    About the business

    A custom software company with 85 employees. At that headcount, routine HR questions and tasks never stop: vacations, health insurance, business trips, and various requests.

    Challenge: Reduce the share of administrative staff.

    Solution: An AI agent with vector search over the knowledge base (RAG) and tools. It answers with a link to the current SharePoint document, files requests in the messenger, and escalates hard cases to HR with dialogue context. To save tokens, keyword rules and messenger buttons cover part of the flows — the LLM is used only where free-form dialogue is needed.

    Delivery and setup time

    3 weeks

    70%
    Tickets handled by AI
    up to 15 seconds
    Reply time to an employee
    up to 2 minutes
    Time to file a request

    Before

    • 1.Many repeat HR questions about vacations, health insurance and business trips.
    • 2.Regulations scattered across SharePoint and file storage.
    • 3.Requests filed manually over email and messenger.

    After

    • 1.The agent answers from the knowledge base with a link to the document.
    • 2.Employees get an answer in seconds, without digging through files.
    • 3.Requests are filed in chat; most tickets close without HR.
    Konstantin M.
    We have 85 people and a large body of internal regulations; a big share of HR tickets used to be the same questions about vacations, voluntary health insurance and business trips. The assistant finds answers in our documents with a link to the source and prepares the request — the HR team only approves. Rollout needed no internal process redesign: the assistant is wired into our corporate messenger and SharePoint.
    Konstantin M. — CEO, IT-Megastar
  • Arasar

    Online building demolition cost calculator

    About the business

    An industrial demolition company. For each request it estimates cost and timeline and prepares a technical and commercial proposal.

    Challenge: Speed up preparation of quotes or rough estimates.

    Solution: An online quote calculator with AI photo analysis of the site. Pricing logic was rebuilt with the estimators from their spreadsheets and coefficients; the model flags hidden risks from photos and returns a stage-by-stage quote. Part of the prompt-based AI logic was exposed in the UI so the client's team can adjust it as they use the tool — closing the feedback loop.

    Delivery and setup time

    8 weeks

    +13%
    Conversion from a fast estimate
    double
    Risk check: expert + AI
    ±15%
    Preliminary estimate vs. final quote deviation

    Before

    • 1.Preparing a quote took 3–7 days in engineers' spare time — some requests went to competitors.
    • 2.Hidden demolition issues were discovered only on site.
    • 3.Different experts assessed the same site differently.

    After

    • 1.A preliminary cost by site parameters goes to the client instantly.
    • 2.AI highlights risks and reduces expert error.
    • 3.The assessment procedure is unified.
    Anton T.
    Technical presale in our industry is heterogeneous: every site differs in area, materials, height, location and access. A rough quote used to take about a week — managers chased engineers, estimators and logistics in their spare time. Now parameters and photos go into the system, and within minutes we get a preliminary quote broken down by stage with risk callouts. Clients hear back the same day; engineers join at final scoping. The Pyramind team digitised our workflows, working through our calculation spreadsheets with our estimators, so the logic matches how we really price. They also researched comparable solutions abroad.
    Anton T. — Development Director, Arasar Group
  • Sociocybernetics

    Argument visualisation for FUKO

    About the business

    The FUKO EdTech platform: participants hold long text discussions that need clear structure.

    Challenge: Turn long discussions into a graph of theses.

    Solution: An LLM text-analysis module with Mermaid diagram output. Not off-the-shelf prompts: we first dug into the FUKO methodology and, following its feedback cycle, split discussion analysis into several stages — from extracting theses to assembling the tree graph. The module sits inside the existing Django platform via API.

    Delivery and setup time

    4 weeks

    minutes instead of hours of manual review
    Discussion processing and diagram build
    Mermaid, SVG, PNG
    Diagram export formats
    within one sprint
    MVP module time-to-launch

    Before

    • 1.No component automatically structured participants' long answers.
    • 2.FUKO terminology had to be absorbed amid open requirements.
    • 3.Needed a coherent discussion diagram, not tables and lists.

    After

    • 1.An LLM pipeline extracts theses and builds a tree.
    • 2.The module plugs into the platform via API: text in, diagram out.
    • 3.Any discussion becomes a graph of theses automatically.
    Mark F. O.
    FUKO is a holistic methodology of cyclic feedback with its own vocabulary and philosophy of dialogue. Honestly, we did not expect an external team to ramp up that fast. Pyramind learned our product from pitch transcripts and by scraping our website. On that basis they designed and embedded a new module in the platform; we iterated together until it matched what we wanted.
    Mark F. O. — Head, Sociocybernetics
  • Technologika

    HR EDM for Technologika

    About the business

    IT outsourcing across several legal entities. HR handles contracts, leave, trips and sick leave for a large staff.

    Challenge: Speed up HR paperwork and remove manual entry into 1C:ZUP.

    Solution: An HR EDM system with e-signatures and automatic sync to 1C:ZUP.

    Delivery and setup time

    1.5 months

    1 day instead of a week
    HR document approval
    fully automated
    Sync with 1C:ZUP
    1.5 months
    Time to launch

    Before

    • 1.HR for several entities ran in Word, Excel and email — documents got lost, approvals dragged on for weeks.
    • 2.Templates were maintained manually; errors surfaced only in audits.
    • 3.Data was moved into 1C:ZUP by hand — mismatches were inevitable.

    After

    • 1.Requests from mobile, templates in the self-service portal.
    • 2.E-signing (UNEP/UKEP) right in the system.
    • 3.Auto-sync with 1C:ZUP — approval in a day.
    Alena K.
    The group includes several legal entities, and the HR team previously spent a large share of the working day routing orders for signature, waiting for scans and re-entering data into 1C. HR specialists now manage templates in the self-service portal while employees submit applications from a mobile device — data flows into 1C automatically. Pyramind quickly delivered an MVP and then continued working with our HR department to refine multiple approval workflows.
    Alena K. — Chief Accountant, Technologika

  • Speech analytics for moving crews

    About the business

    A freight and moving service: crews complete hundreds of jobs a month without a supervisor on site.

    Challenge: Control crew quality and integrity without a supervisor present.

    Solution: Shift speech analytics: recording, transcription and AI review with anti-fraud. Before analysis we segment the recording and keep only client-interaction fragments — saving tokens on long shifts. The model catches conflicts, script misses and off-the-books payment attempts.

    Delivery and setup time

    5 weeks

    100%
    Shifts under control
    0%
    Off-the-books payments
    same day as the shift
    Dispute review

    Before

    • 1.Hundreds of jobs without a supervisor — what happens on site is opaque.
    • 2.Some orders were paid off the books.
    • 3.Complaints came after the fact, with no evidence.

    After

    • 1.Every shift is recorded and reviewed by AI: scripts and tone monitored.
    • 2.Audio capture ended off-the-books payments; isolated cases are caught immediately.
    • 3.Disputes are reviewed on the day of the shift.
    We run hundreds of jobs a month, and crews work on site without a supervisor — we simply could not assess what actually happened during a move. Complaints surfaced after the fact with no evidence: the client's word against the movers'. Now every shift is recorded and reviewed automatically: the system checks the greeting and upsells, flags conflicts and raised voices, and separately marks attempts to arrange payment directly, off the books. We see disputes on the day of the shift, not a week later. The Pyramind team understood the specifics of field work: they tuned recognition for street noise and crew speech, and we refined the event tagging together on real recordings.
    Operations Director — Freight transport and moving service (under NDA)

Want a similar outcome?

Tell us about your task — we will sketch a concept and a subscription estimate.

Discuss your projectCalculate subscription

Cookie preferences

Legal documents

Privacy Policy

Privacy Policy

Effective date: June 10, 2026

This Privacy Policy explains how Pyramind, a sole proprietorship (preduzetnik) registered in the Republic of Serbia (the “Operator”, “we”, “us”), processes personal data of visitors and prospective clients of the website https://pyramind.tech (the “Site”) and of users of products developed by Pyramind under separate subscription agreements.

This document is drafted to comply with the Serbian Law on Personal Data Protection (Zakon o zaštiti podataka o ličnosti, “ZZPL”) and to align with Regulation (EU) 2016/679 (GDPR), the UK GDPR and DPA 2018, the Brazilian LGPD (Law No. 13.709/2018), the Swiss revised FADP, and the California CPRA to the extent applicable to our processing.

A separate Russian-language policy (/legal/politika-obrabotki-pd.md) governs processing performed by our Russian affiliate (Sole Proprietor Dmitrii Litvinenko, RU) for Russian-resident clients under a separate contractual setup.

1. Controller and contact

  • Operator (controller): Pyramind, sole proprietorship (preduzetnik) registered in the Business Registers Agency of the Republic of Serbia (APR).
  • Registered name: [REGISTERED NAME — to insert]
  • PIB (tax ID) / matični broj (registration number): [to insert]
  • Registered seat: [address], Republic of Serbia
  • Email for privacy enquiries: dev@pyramind.tech
  • EU/UK Article 27 representative: to be appointed in writing before active marketing in the EU/UK and listed here. Until appointment, please direct all enquiries to dev@pyramind.tech.
  • Data Protection Officer: not appointed; under GDPR Art. 37 / ZZPL Art. 56 not strictly required at the current scale of processing. We will reassess this if processing volume or sensitivity changes.

2. Personal data we process

We process the following categories of personal data:

  • Contact data you submit through forms or by emailing us: name, business email, phone number, company name, company URL, preferred contact method (call / Telegram / WhatsApp), free-form message.
  • Technical data automatically collected when you browse the Site: IP address, user-agent, referrer, time-zone, language, device and browser characteristics, pages viewed, events triggered. This data is collected only after analytics consent for non-essential cookies (see Section 6).
  • Communication data: emails, messages, transcripts of voice or video calls if you initiate them. We record calls only with your prior explicit consent.
  • Account / contractual data for clients under a Subscription Agreement: legal entity details, signatory names, billing data, invoice records.

We do not intentionally collect special categories of data (Art. 9 GDPR / Art. 17 ZZPL), data of children under 16 (Art. 8 GDPR), or biometric data via the Site. If you submit such data voluntarily through a free-form field, we will minimise and erase it.

3. Purposes and lawful bases

PurposeLawful basis (GDPR Art. 6 / ZZPL Art. 12)
Responding to enquiries, sending proposals, conducting pre-contractual negotiations(b) steps prior to entering into a contract; (f) legitimate interest in B2B outreach
Performance of a Subscription Agreement and related services(b)
Site analytics, performance monitoring, A/B testing(a) consent (granular, withdrawable); (f) for strictly necessary security telemetry
Marketing newsletters and re-engagement(a) opt-in consent (no soft opt-in for non-customers)
Compliance with tax, accounting, AML, sanctions screening, court orders(c) legal obligation; (f) legitimate interest in fraud and sanctions risk control
Defending legal claims(f); (e) under Serbian public-interest grounds where applicable

For California residents, the corresponding business purposes under the CPRA are: providing services, performing contracts, security and integrity, debugging, internal research, and limited marketing.

4. Recipients and processors

We share personal data with the following categories of recipients:

  • Hosting and infrastructure providers — by default within the European Economic Area or the Republic of Serbia. For products delivered to clients in other regions, we may, with the client's prior agreement, use a regional cloud (UAE, Singapore, etc.).
  • Email and messaging providers for replying to your enquiries (transactional email, Telegram, WhatsApp Business).
  • Analytics providers — Yandex LLC (Yandex Metrica), only if you have given analytics consent. Yandex Metrica processes data in the Russian Federation; see Section 5.
  • Payment, banking, and accounting service providers in the Republic of Serbia and, where applicable, in the recipient's country.
  • Professional advisors — lawyers, auditors, tax advisors — under written confidentiality.
  • Authorities — when required by Serbian law, EU/EEA law applicable through SCCs, court order, or to protect our rights.

A current list of our sub-processors for product clients is published as a separate document; see `/legal/en/sub-processors.md`.

5. International data transfers

The Operator is established in the Republic of Serbia. Serbia is not an EU member state and is not subject to a current adequacy decision under GDPR Art. 45. However, the Serbian Commissioner for Personal Data Protection has confirmed Serbian DP law as substantially aligned with the GDPR, and Serbia is on the Council of Europe Convention 108+.

For transfers from the EEA, UK, or Switzerland to Serbia we rely on:

  • Standard Contractual Clauses (SCCs) under Commission Implementing Decision (EU) 2021/914 with supplementary measures (encryption in transit and at rest, access controls, audit logging) for ongoing engagements;
  • Your explicit, informed consent under Art. 49(1)(a) GDPR for one-off interactions where appropriate;
  • Contractual necessity under Art. 49(1)(b) where the transfer is necessary for the performance of, or the conclusion of, a contract you requested.

Where analytics consent is granted, technical data may be transmitted to Yandex (Russian Federation). Russia is not subject to an adequacy decision; we rely on your explicit consent (GDPR Art. 49(1)(a)) for this transfer, with the disclosure that Russian law may grant authorities access to data without remedies equivalent to EU standards.

6. Cookies and similar technologies

The Site uses cookies and similar technologies in three categories: strictly necessary, analytics, and marketing. Analytics and marketing cookies do not load until you give consent through our cookie banner (all non-essential storage defaults to denied). Detailed information on each cookie, retention periods, and opt-out controls is set out in our Cookie Policy.

You can withdraw or change cookie consent at any time via the “Manage cookies” link in the footer or on this page.

7. Retention

We retain personal data only for as long as necessary for the purpose of processing, then delete or anonymise it.

  • Enquiry contacts (no contract concluded): up to 12 months from last contact, then deleted.
  • Client contractual records: for the duration of the Subscription Agreement and 3 years after termination, plus the longer of (i) the statutory accounting and tax retention period under Serbian law (5 years for primary accounting books, 10 years for invoicing records, per the Serbian Accounting Law) and (ii) any retention required under your local law.
  • Analytics data: as configured in the analytics provider (default 25 months for Yandex Metrica, unless you withdraw consent earlier).
  • Marketing consents: until you withdraw consent or for 3 years from last engagement, whichever is earlier.

8. Your rights

Subject to applicable law, you have the right to:

  • Access the personal data we hold about you (GDPR Art. 15 / ZZPL Art. 26 / UK GDPR / LGPD Art. 18 / CPRA §1798.110);
  • Rectify inaccurate or incomplete data (Art. 16 / ZZPL Art. 29);
  • Erase data (“right to be forgotten”) (Art. 17 / ZZPL Art. 30);
  • Restrict processing (Art. 18 / ZZPL Art. 31);
  • Data portability in a structured, commonly used, machine-readable format (Art. 20 / ZZPL Art. 36);
  • Object to processing based on legitimate interests, including profiling (Art. 21 / ZZPL Art. 37), and, in particular, to direct marketing at any time;
  • Withdraw consent at any time, without affecting prior lawful processing (Art. 7(3) / ZZPL Art. 15);
  • Not be subject to a decision based solely on automated processing producing legal effects (Art. 22 / ZZPL Art. 38). We do not use such processing.
  • For California residents: the right to know, delete, correct, opt out of sale or sharing, and limit the use of sensitive personal information (we do not “sell” personal data within the meaning of CPRA §1798.140(ad), but we will honour Global Privacy Control (GPC) signals).
  • For Brazilian data subjects: the rights enumerated in LGPD Art. 18.

To exercise rights, email dev@pyramind.tech. We will respond within 30 days (GDPR/UK/LGPD/ZZPL) or 45 days (CPRA), with one possible extension as permitted by law. We may need to verify your identity before disclosing or modifying data.

You also have the right to lodge a complaint with a supervisory authority — the Serbian Commissioner for Information of Public Importance and Personal Data Protection (Poverenik za informacije od javnog značaja i zaštitu podataka o ličnosti) for matters under our remit; your local data protection authority in the EEA / UK; the ANPD in Brazil; or the California Attorney General / CPPA.

9. Security

We implement organisational and technical measures appropriate to the risk under Art. 32 GDPR / Art. 42 ZZPL, including encryption of data in transit, access controls, audit logging, regular backups with restoration testing, vulnerability management, and incident response procedures. In the event of a personal data breach affecting your data, we will notify the Serbian Commissioner within 72 hours where required, and notify you without undue delay where required by law.

10. Automated decision-making and profiling

We do not engage in automated decision-making producing legal or similarly significant effects on you (GDPR Art. 22 / ZZPL Art. 38). Where products we develop for clients embed AI/ML features, the client (controller) is responsible for compliance with Art. 22, the EU AI Act (Regulation (EU) 2024/1689), and equivalent provisions; we act as processor under a separate Data Processing Agreement.

11. Children

The Site is not directed to children under 16. We do not knowingly collect personal data from children. If you believe we have collected such data, please contact us and we will delete it.

12. Changes to this Policy

We will publish updates to this Policy on this page with a new effective date. Material changes will be highlighted at the top of the page for at least 30 days, and we will notify clients with a current Subscription Agreement by email.

13. Contact

For all privacy-related enquiries, requests, or complaints:

  • Email: dev@pyramind.tech
  • Postal address: [address], Republic of Serbia (available on request from the email above for service of formal legal correspondence).

This Privacy Policy is provided for information only and does not, by itself, replace contractual or statutory protections. In case of conflict between this English text and any local-language version, the English text governs for all data subjects outside the Russian Federation; for Russian-resident data subjects whose data is processed by our Russian affiliate, the Russian-language /legal/politika-obrabotki-pd.md prevails.