Ahmad Rasoulpour

• Driving the AI document-intelligence feature on SimpleCitizen, Fragomen's consumer immigration platform: an LLM extraction pipeline reads users' uploaded documents (passports, visas, civil records) and auto-fills their immigration questionnaire and application forms - automating ~80% of form-filling and cutting typical 4-5-hour application prep by roughly 80%, with user review of every AI-filled field.
• Designed and implemented a full-stack in-platform service-purchase flow (Angular, TypeScript, GraphQL, Django), leading a team of three engineers; drove a 15% revenue increase within four months of launch.
• Directed the payments migration from Authorize.Net to Stripe, supporting a 50% increase in transaction volume.
• Led a team of four building a modular frontend component architecture serving 200,000+ monthly active users.

• Spearheaded lazy loading on core routes, cutting initial load times 20-50% and directly improving engagement and retention.
• Optimized GraphQL queries and caching, decreasing data-fetching time by 40% and server load by 25%.
• Owned diagnosis and resolution of production incidents, prioritizing long-term maintainability over quick patches.

• Designed and implemented a new B2B dashboard for partner immigration case management, including quota distribution across partner teams and subcontractors (Angular, Django, REST).
• Spearheaded the platform migration from REST to GraphQL, reducing response times by 40%.
• Built a reusable Angular web-component system that improved internal workflow efficiency by 40%.

• Built a production real-time AI phone-call agent: Claude generates a professional English call script from a parent's request written in any of 11 languages, then a Vapi-orchestrated pipeline (Deepgram speech-to-text to Claude to ElevenLabs text-to-speech) places calls to insurance, Medicare, and government offices, with webhook-driven transcripts summarized bilingually for the family.
• Shipped AI document intelligence (Claude vision: classification, translation, four-level urgency triage, plain-language action steps), scam detection with structured risk verdicts, and a live two-way translator - each responding in the parent's native language, including RTL scripts.
• Designed a FastAPI backend with 92 endpoints: dual-principal JWT auth (phone-OTP for parents, OAuth for family members), rate limiting, and four inbound webhook integrations (Stripe, Vapi, RevenueCat, Twilio).
• Cut LLM inference costs with model tiering (Claude Sonnet for reasoning, Haiku for translation/classification) and prompt caching; built metered usage billing (Stripe + RevenueCat) with per-tier call-minute allotments and one-time credit packs.
• ~21K lines solo across a Next.js web app, Expo/React Native mobile app (34 screens, 11 locales), Python backend, and Telegram/WhatsApp channels.

• Built a streaming AI tutor with a provider-agnostic backend (Anthropic SDK plus any OpenAI-compatible API via hand-rolled SSE parsing), with context-aware system prompts injected per lesson.
• Implemented in-browser Python execution with Pyodide/WASM (numpy/pandas auto-loading), an FSRS spaced-repetition review scheduler, and a ReactFlow skill tree with prerequisite locking over an 89-lesson typed curriculum.
• Shipped the full SaaS loop live at takebelay.com: email + Google auth, Stripe subscriptions with a lifetime tier, dual-driver database layer (Neon Postgres / SQLite), Dockerized production deployment on Fly.io.

• Co-built a complete iOS/Android game (team of two): ~52K lines of C# across 180 scripts (Unity 2022 LTS) spanning a summon-place-merge combat loop, a weighted-probability gacha engine with soft-floor odds and board-state success decay, a multi-currency economy, battle pass, and 35 custom editor tools that procedurally bake UI prefabs and scenes.
• Wrote a Supabase cloud-save backend from scratch in C# over UnityWebRequest (PostgREST upserts, JWT auth, deep-link OAuth, debounced offline-tolerant sync); integrated Unity IAP subscriptions and GDPR-compliant rewarded ads.

Also built: GameForge (generative game-asset pipeline: Llama 3.3 JSON-mode stats, SDXL LoRA sprites, Stable Audio SFX, Unity-ready packaging), an algorithmic BTC trading system (walk-forward backtester modeling fees and slippage, exchange-side stops, VPS deployment), and a multi-chain DeFi transaction classifier with Form 8949 CSV export.