EN
Download CV

Founder & CEO, A Generator · Tashkent

AleksandrChernikov

A Generator is a scheduling engine sold as an API. We start with schools: school software calls our API to build the timetable and find substitutes; schools without such software use Vyorm Schedule, our web app on the same engine. The engine has new best-known results on 4 of 25 instances of XHSTT-2014, the international school timetabling benchmark, and 4 new best-known solutions in nurse rostering. I write all of the code, with Claude Code. I also run Alrevion, my product studio in Tashkent.

Message on Telegram Download CV (PDF)
new best-known results on XHSTT-2014 (school timetabling)
4
new best-known solutions in nurse rostering
4
products built at Alrevion, my studio
12
launched to real users
5

Experience

A Generator is my company. Alrevion is my product studio in Tashkent.

Sep 2026 — nowTashkent

Founder & CEO, A Generator

A scheduling engine sold as an API; I write all of the code, with Claude Code.

  • Engine. Our own Rust core of about 75,000 lines, written from scratch. For a real school (1,711 lessons in its weekly timetable) it builds a complete clash-free timetable in 3.5 s once the data is in; it measures how close each timetable is to the best possible and explains when a task has no solution.
  • Product. The API has been live since 29 October 2026: school software calls it to build the timetable and find substitutes. Schools without such software use Vyorm Schedule, our web app on the same engine (a demo in testing).
  • Benchmarks. New best-known results on 4 of 25 XHSTT-2014 instances, verified with the official HSEval evaluator, submitted to the XHSTT archive, not yet published; 4 new best-known solutions and 6 new lower bounds in nurse rostering, accepted by schedulingbenchmarks.org (details below).
  • Roadmap. Schools, then universities, then hospital and shift rostering, then vehicle routing.
2026 — nowTashkent

Founder, Alrevion

My product studio in Tashkent: 12 products built, 5 launched to real users; fixed-price and hourly work for local clients.

  • Shared platform. Alrevion Auth (OTP, password + TOTP 2FA, passkeys / WebAuthn, social sign-in; JWT / JWKS) and Alrevion Push (Web Push, APNs, FCM; scheduling, retries, signed webhooks) — one identity and notification layer for every product, edge-native on Cloudflare Workers.
  • Delivery. Specification and plan before code (domain model, constraint catalog, ADRs), weekly working increments, acceptance through independent verification; clients assemble the scope in an interactive estimator with a live price and timeline (three-point estimates with risk reserves); AI-native workflow with coding agents and adversarial review; pre-seed pitch deck and financial model.
2024 — 2026Project-based

Independent Business Consultant

Freelance · manufacturing and services SMEs

  • Restructured the sales funnel at a company that manufactures refrigeration equipment; pricing and targeting changes led to an increase in monthly revenue of at least 20% within 3 months.
  • Implemented a CRM system and task automation at a service company; this reduced manual work and cut client response times by about a third.
  • Modeled AS-IS processes, found automation opportunities and prepared requirements and a digital-infrastructure roadmap for owners and managers.

A Generator

My company. The engine builds the timetable and measures its quality.

A scheduling engine sold as an API

ag.alrevion.com

Lessons, rooms, teacher workloads and health rules go in; a clash-free timetable with a measured quality score comes out. We start with schools: school software calls our API to build the timetable and find substitutes, and schools without such software use Vyorm Schedule. Next come universities, then hospital and shift rostering, then vehicle routing. The engine runs on our own Rust core, about 75,000 lines written from scratch: it measures how close each timetable is to the best possible and explains when a task has no solution.

School timetabling: new best-known results on 4 of 25 XHSTT-2014 instances
InstancePrevious bestA GeneratorImprovement
NL-KP-091,620665−59%
NL-KP-03199102−49%
DK-FG-121,2631,020−19%
NL-KP-05425399−6%

Soft-constraint penalty, lower is better; 0 hard-constraint violations. Previous bests: GOAL, 17 April 2017 (XHSTT archive). Our solutions are verified with the official HSEval evaluator and were submitted to the XHSTT archive on 14 and 28 October 2026; not yet published. Our scores match HSEval on all of the benchmark’s 25 problems from real schools in 12 countries. Solution files are available on request.

Nurse rostering: new best-known solutions, accepted by schedulingbenchmarks.org (21 and 27 October 2026)
InstancePrevious bestA GeneratorImprovement
MER-A8,7798,328−5.1%
ERRVH-A2,1182,076−2.0%
ERRVH-B3,0903,027−2.0%
CHILD-A21,0871,086−0.1%

Plus 6 new lower bounds, accepted on 27 October 2026; all on the site’s original nurse rostering instances, listed on the benchmark changes page. Our reading: for ERMGH-A and ERMGH-B the new lower bounds meet the best-known solutions, so those are optimal.

XHSTT-2014 problems where our scores match HSEval
25 / 25
countries in the benchmark: 25 problems from real schools
12
to build a complete clash-free week for a real school (1,711 lessons)
3.5 s
API in production since 29 October 2026
Live

Idea → Delivery

How I take a product from an idea to a working system.

From idea to MVP and delivery

my own method

This is how I run products myself. Every stage ends with a question, an artifact and a decision: continue, stop, hold, or change the hypothesis. The rule is to learn cheaply whether to go on — a conversation first, then a prototype, then a prepayment, and only then code. I also wrote the method down as a skill for AI agents, so the same process and templates run on every project.

New product

the idea is not validated yet
  1. SIdea shaping, USP, JTBD
  2. 0Hypothesis
  3. 1Customer interviews
  4. 2Competitors & market
  5. 3Prototype
  6. 4Paid experiment
  7. 5Unit economics (CAC, LTV)
  8. 6Requirements
  9. 7MVP
  10. 8Pilot
  11. 9Decision

Client project

a client brings a task
  1. C0Charter
  2. C1Survey
  3. C2Business case (NPV, IRR, payback)
  4. C3AS-IS / TO-BE processes
  5. C4UX
  6. C5System analysis & specification
  7. C6Pilot
  8. C7Delivery & acceptance
  9. C8Launch, SLA, change management
tracks with a hand-off between them
2
stages, each with a go / kill gate
20
document templates (charter, spec, RACI, SLA…)
16
financial checker: NPV · IRR · PI · DPP · CAC · LTV
1

Products

Led at Alrevion, from specification to production.

In productionMarketplace

IJODA

Uzbekistan’s first marketplace of verified creators — photographers, videographers, bloggers and models. Clients find, compare and book them in one place.

ijoda.alrevion.com
In productionEvents

Tribeon

Hackathons and olympiads on the web and in Telegram: sign-up, team matching by role, check-in and certificates. First hackathon: 54 sign-ups, 47 checked in, 16 teams.

tribeon.alrevion.com
In productionPlatform

Alrevion Auth

One API for one-time codes, password + 2FA, passkeys and social sign-in. Products verify tokens locally via JWKS.

auth.alrevion.com
In productionPlatform

Alrevion Push

Web Push, APNs and FCM through one API: per-device routing, retries with backoff, scheduled sends, signed webhooks.

push.alrevion.com

Also: tennis school SaaS (client, used in production until July 2026); in development — Ready EIP (corporate learning with an AI role-play trainer), Vyorm Schedule (school timetabling web app on A Generator; demo), Alrevion PMS (short-term rentals), Hando (voice AI assistant in Telegram); paused MVPs — Stomiko (software for solo dentists), Family Finance AI (hackathon).

Education & awards

2022 — 2026St. Petersburg

B.Sc. Business Informatics — E-Business

Peter the Great St. Petersburg Polytechnic University (SPbPU)

GPA 4.35 / 5.0, credit-weighted (≈ 3.35 / 4.0 US)Admitted 13th of 193 to a fully state-funded place
  • Thesis: an intelligent information system for scheduling and the educational process, piloted at a private school in Tashkent before the defense.
  • Coursework: Enterprise Architecture, Business Process Modeling, Data Management, Digital Marketing & E-Commerce, IT Project Management.
2024 — 2025256 hours

Professional retraining diploma — IT Consulting

SPbPU · qualification: Information Systems Manager (ICT)

  • Business process optimization, enterprise architecture, IT project management methodologies, digital transformation and information security, industry cases, IT consulting practice.
  • Final project (graded excellent): a customer-support chatbot that automates request handling.
2023 — 2026Competitions
  • 1st in the programming track, “IT Perspective” (Rostelecom IT × SPbPU, 2025), one of 5 track winners among 60+ teams; recommendation letter from Rostelecom’s Director of Strategic Projects Development.
  • Top-20 — Avito BootCamp, real company case (March 2025).
  • Family Finance AI — Cursor Hackathon (April 2026): working MVP that turns bank transaction history into a household financial plan with an AI assistant.
  • CMS & corporate website — full business-analysis project (December 2025): SWOT, BMC, multi-criteria CMS selection, technical specification, test plan, UAT, CAPEX/OPEX and ROI, MS Project schedule.
  • UniHaven — Startup Weekend × ITMO (November 2023): rental marketplace MVP and business model in a weekend.

Skills

Product & strategy

Product discovery · Customer interviews (JTBD) · Competitor teardowns · Value proposition & USP · Demand tests (landing page, fake-door pricing) · Usability tests on clickable prototypes · MVP definition · Roadmapping · Go-to-market · Business model · Product analytics (funnels, retention, A/B tests) · Subscription metrics

Business & systems analysis

Requirements engineering (functional / non-functional, acceptance criteria) · Process modeling L0–L3, AS-IS / TO-BE (BPMN, EPC, IDEF0) · Business rules · Domain & data modeling (conceptual, logical, physical) · Technical specification · System architecture & ADRs · API integration · Roles & access rights · Threat modeling for personal and medical data · UAT & acceptance

Project management

Agile delivery · Sprint planning · Risk & stakeholder management (RACI) · Escalation matrix L0–L3 · Spec-first delivery · ITSM, SLA & runbooks · Role guides · Change management · MS Project · Notion · Figma

Finance

Financial modeling · Business case (NPV, IRR, PI, payback) · ROI · CAPEX / OPEX · Unit economics

Estimation & commercial

Three-point (PERT) estimates · QA, PM and risk reserves · Commercial proposals · Interactive project estimators for clients · Payment models: staged, revenue share, support retainers · Build vs buy vs subscribe · Legal requirements checklists (offer, privacy policy, consents)

Optimization

Combinatorial optimization · Lower bounds · Benchmarking (XHSTT-2014 with HSEval; nurse rostering)

Engineering

TypeScript · Python · Rust · SQL · React · React Native · Vue · Astro · Hono · PostgreSQL · Telegram bots & Mini Apps · Payment integrations (Click, Payme) · Cloudflare Workers, D1, R2 · Google Cloud Run

Security

Passwordless auth · WebAuthn / passkeys · OTP & TOTP 2FA · JWT / JWKS · Rate limiting · PII hashing

AI

LLM product integration · LLM evaluation (quality, safety, latency, cost per dialogue) · AI coding agents · Writing agent skills · Speech-to-text · Prompt design

Languages

Russian — native · English — C1 (EnglishScore)

Certificates

Anthropic — Claude 101, AI Fluency Framework, AI Fluency for Nonprofits (2026) · HubSpot — Digital Advertising (2026) · EnglishScore C1 (2026)

Contact

The fastest way to reach me is Telegram or phone.

Telegram +998 97 029-11-22 [email protected]
© 2026 Aleksandr Chernikov Tashkent · alrevion.com