Hi. I'm Sukhrob Tokhirov.

• Built the tariff engine at the core of street-parking billing, co-designing its rule logic with another developer and the systems analysis lead. Each zone gets its own tariff of prioritized rules with AND/OR conditions (time windows, subscription type, resident status, and more), grace periods, and a base-price fallback. It prices sessions for ~150K monthly users and ~1M transactions a month.
• Built the gated-parking camera pipeline: a Linux service wrapping the Dahua Java SDK and DSS that connects to multiple ANPR cameras and publishes to Kafka. One stream carries device status, so faults alert that zone’s staff right away; the other carries vehicle events, logged raw and turned into parking sessions automatically. Live session state is kept in Redis.
• Prepared the training dataset for a YOLO-based plate recognizer built with a colleague, and took it to production at ~98.7% accuracy across 39K real reads (~50 ms per recognition, under 250 ms end to end).
• Owned auto-renewal and auto-charging for sessions and subscriptions across saved cards and wallet balance. Built card payments with Click, Payme, and Octo alongside the lead developer.
• Led the large external integrations: Flexy (insurance providers), Yandex Plus and Yandex Maps, DriveLense (vehicle photos and video), and the advertising and public-offers service, which I own.
• Worked on the core session logic during the PHP-to-Java migration. Changed 15+ of the platform’s 30+ services and added integration tests across all of them, plus unit tests where coverage was missing.
• Coached the 10-person team on using AI coding tools well.

• Building a self-hosted AI server running Qwen3.8-27B, with a vector database, RAG, and structured-extraction pipelines, so company data stays in-house instead of going to a third-party API.
• Researching the cost of self-hosting against hosted APIs to keep early infrastructure spend low while usage is still small.
• Building the backend on Java 25 and Spring Boot 4 with virtual threads, with real-time features over WebSockets, Redis, Kafka, and LiveKit, and an Angular frontend.
• Integrated Google Calendar, Google Docs, and Telegram bots.

• Author verifiable coding tasks used to evaluate AI coding agents on real open-source repositories, mostly in Python and C++, as part of Shipd’s Olympus program.
• Research live codebases to find unimplemented features, then write the specification, a hermetic containerized test suite, and a reference implementation for each.
• Design tests strict enough to block shortcuts without pinning a single implementation, so correct alternative solutions still pass. Keep task descriptions free of solution hints and calibrate difficulty against live agent runs.

Ride-hailing platform (Uber/Bolt-style) built for the Romanian market; pre-launch, currently in market selection. One of 3 backend engineers.

• Built driver matching on H3 hexagonal zones, using PostGIS for geospatial queries and Redis Pub/Sub plus WebSockets to stream live driver GPS.
• Built a Redis-backed surge-pricing system that reprices geozones in near real time based on supply and demand.
• Built rider payments and driver payouts on Stripe Connect (separate charges and transfers, with holds for dynamic fares), plus Apple Pay and Google Pay through Stripe.
• Built the stateless auth layer (RSA-signed JWT, Redis sessions, OTP); load-tested with JMeter at 3,000 req/s with p99 logins under 500 ms.
• Designed the data layer for GDPR, keeping each user’s data on servers in the correct region.
• Deployed 30+ services to AWS EKS with zero-downtime rollouts. Co-wrote the Terraform with the lead developer: I provisioned the application services and servers; he set up the database, Redis, and RabbitMQ. Most services have ~90% unit-test coverage.
• Built async event flows over RabbitMQ, and push notifications through FCM and APNs, including iOS Live Activities for ride status.

• When Meta announced it was shutting down WhatsApp Lite, built a dedicated WhatsApp Cloud API section in Kommo’s main navigation, with a full setup and configuration flow, to move customers to Cloud API before the shutdown and protect revenue.
• Key engineer on a 5-person team: worked with the lead developers and product owners on technical decisions, then ran the implementation-planning meetings.
• Integrated NVoip audio and video calling. Calls are recorded to Google Cloud Storage and attached to the task’s chat history, where users can replay and download them.
• Built in-chat Stripe invoicing: business owners create and send invoices inside the conversation, and every payment event is recorded in the chat history.
• Fixed a rare Google Sheets sync bug affecting a single client whose device-details data exceeded the 512-character column limit, by splitting oversized values across multiple rows.
• Processed incoming integration webhooks, such as Stripe events, asynchronously over RabbitMQ with retries.

Stack: PHP (Symfony, Laminas), REST, gRPC, WebSockets, MySQL, Redis, RabbitMQ, Kubernetes, GitLab CI

• Built and maintained internal platforms for the PDP ecosystem (the pdp.uz online programming school, PDP University, and PDP School), covering their ERP, LMS, and CRM systems.
• Developed features across 30+ Spring Boot and Spring Cloud microservices (Eureka, OpenFeign), with PostgreSQL, Liquibase migrations, and Redis caching.
• Integrated Hikvision face-recognition terminals for staff attendance and student check-in. This project, together with bug fixes during my trial period, turned the internship into a full-time role.