Ilnar Mukhutdinov

US-based AI procurement platform used by enterprise sourcing teams. Responsibilities: - Backend development of identity and access management (IAM) services - Design and delivery of event-driven services on Kafka - Securing AI coding agent tooling and its access to internal systems - PostgreSQL schema and query performance work - Cloud infrastructure across AWS and Azure Key results: - Designed and delivered the in-app notification service on Kafka, with topics partitioned by tenant id so tenant-wide events reach users in order, and per-user inboxes in PostgreSQL - Delivered a TOTP-based MFA system integrated into the existing auth pipeline, which unblocked contracts with enterprise customers that required MFA - Built SCIM role provisioning for internal tooling and enterprise customer onboarding - Designed and shipped a token exchange service in Rust (Axum) handling ~1k RPS at under 3ms latency, which removed HTTP header size limits and generalized the auth pipeline for internal IAM flows - Built a spec-driven workflow for the coding agent, with guardrail hooks and a test gate with a retry budget - Secured the coding agent's MCP server with scoped, short-lived sessions and human approval for writes - Migrated services from AWS to Azure

Fairmarkit

Cyprus-based analytics startup measuring audience emotional responses to video creatives and gameplay through facial expression analysis. Responsibilities: - Technical leadership of the analytics platform and its engineering team - Design and development of backend infrastructure for user management, reporting, billing, and data processing - Maintaining GitLab CI/CD and Terraform state for all production microservices Key results: - Hired and led a team of six engineers across backend, frontend, QA, and DevOps - Built a full-stack analytics platform processing over 2.5M emotional response records monthly from more than 10K research participants daily - Designed the data layer with hot/cold tiering: partitioned ClickHouse for raw emotion values, GCS for cold storage - Built queue-based processing pipelines on Celery and RabbitMQ that scale horizontally by adding workers - Led the split of the monolith into microservices supporting the video processing pipeline - Built a ledger-style payment system on Stripe (card and bank transfer) that records each balance change as an entry to handle concurrent updates, and worked with legal counsel to encode tax rules for Germany, Cyprus, and the USA - Designed a service for video interview recording that let the internal team capture gameplay sessions - Integrated AI-based insight generation for video creatives, cutting analysis time by ~50%

Emhance (formerly Sensemitter)

Emotion analytics company measuring viewer reactions through facial expression analysis and gaze tracking. Responsibilities: - Backend development of microservices for the emotion analytics pipeline - Refactoring existing code and adding test coverage - Monitoring system integration Key results: - Designed and built the emotion detection and gaze tracking microservice, extracting it from a monolithic Django application; it ran on up to 60 workers and was designed to scale to hundreds - Built a microservice automating heatmap generation, cutting report creation time from around 2 hours to 40 minutes - Built custom OpenVINO inference for face detection, reaching ~200 FPS - Integrated Amazon Mechanical Turk for participant recruitment

Sensemachine

Astronomical research system recognizing celestial objects in infrared imagery, running on Texas Instruments DSPs fed by an Altera Cyclone FPGA. Responsibilities: - Development and optimization of real-time image processing software for DSPs (digital signal processors) - Development of communication protocols and internal tooling for DSP control Key results: - Sped up real-time celestial object recognition by roughly 10x through an Assembly implementation - Built a C#/WPF remote control application for DSP control and infrared frame analysis - Implemented UDP-based protocols for remote DSP control - Implemented frame caching that moves images from DDR memory into CPU cache

Research institute