Eraj Saliev

A live leaderboard experience with real-time ranking, branded competitions, and an editor mode to tune visuals per affiliate.

A production data warehouse built on BigQuery's medallion architecture — real‑time CDC via Datastream continuously mirrors Postgres into BigQuery, parameterised SQL scripts run every 3 minutes on Kubernetes via Python to transform raw data into curated domain tables and consumption aggregates, and a custom Next.js monitoring dashboard (Prefect‑style) tracks pipeline health with Slack alerting on failures.

Three AI agents replace the human analyst. An Analyst with ~10 function‑call tools investigates the data, a Senior Analyst reviews and sends it back for drill‑downs, and a Reporting Agent distills all research into the executive briefing that ships to C‑level inboxes untouched.

A truly agentic AI system with 4 autonomous agents — each with its own Gemini model, system instruction, and structured function‑calling schema. An orchestrator sequences data collection, delegates to specialised agents (Decision → Content → Banner Brief → Image Generation), merges outputs, and routes through Slack for human approval before any customer touchpoint.

An 8‑pass, config‑driven bonus engine that calculates individualised player bonuses from dual surplus pools, applies stacking performance multipliers against tenure‑adjusted baselines, and distributes payouts via Postgres and CleverTap — all without a single hardcoded threshold.

A full‑stack SaaS product that generates tailored resumes and cover letters from any job posting. Users paste a URL, the system scrapes and extracts requirements via a multi‑source fallback chain, then Gemini with extended thinking customises content to match — two‑phase async processing with a ChatGPT MCP integration.

A production microservice that aggregates customer financial statistics from BigQuery with multi‑level caching and concurrent async queries — delivering sub‑10ms cached responses from a 3‑second cold path.

A progression engine that calculates player XP from multi‑source betting data, assigns levels via binary search, and atomically awards three tiers of bonuses — with delta processing so only today's active players are touched.

A streaming CDC alerting system that subscribes to Pub/Sub event streams, monitors database tables with configurable alert rules and thresholds, enriches events via a dual‑layer Redis + BigQuery cache, groups related alerts into Slack threads, and self‑monitors its own health — all with exactly‑once delivery.

A full‑stack Next.js application built end‑to‑end on Azure — Azure AD SSO with guest accounts, Blob Storage for assets, Container Registry + App Service for deployment, and PostgreSQL Flexible Server for data. Features a cut optimisation algorithm (bin packing) that evaluates bar combinations to minimise material waste, and a multi‑tenant architecture isolating product catalogues, pricing, and branding per client.

Built a production data warehouse from scratch on Snowflake for Entain's AU/NZ brands. Azure Container Instances consume high‑volume Kafka streams (bets, deposits, registrations, KYC) via consumer groups, land data in a 3‑layer medallion architecture, and 30+ Prefect flows transform it for Power BI dashboards, ML models, Salesforce CRM, and internal personalisation APIs.

A dual‑mode promotional engine that allocates bet‑back allowances for horse racing events across Entain's AU/NZ brands. A batch cron job runs a Snowflake eligibility matrix to catch all missed allocations, while a real‑time Flask service gives new clients instant allowances via Redis‑cached race data — both routing through the Internal APIs for allocation and SFMC for marketing sync.

A full‑stack Next.js promotional platform built from scratch that lets horse racing owners register their runners with Ladbrokes. The system auto‑verifies ownership against racing authority data in real time, routes verified entries to SFMC and internal APIs, and flags unverified submissions to Slack for manual review — replacing an entirely manual verification workflow.

A real‑time personalisation engine that evaluates every bet settlement against each client's profile — tags, segments, ratings, betting history — to deliver individually targeted promotions. Kafka streams feed dual workers for single bets and parlays, while a Next.js admin portal lets marketing configure campaign rules without code.