Arya

AI and data automation professional with a scientific academic background in Biomedical Sciences and Pharmaceutical Sciences, and hands-on experience architecting, building and operating production AI-enabled platforms.

Experience spans generative AI, LLM and multimodal orchestration, AI agents, PostgreSQL/Supabase data systems, API integrations, automated data pipelines, analytics, testing, cloud AI and cost-controlled inference. Built production systems that transform unstructured information into structured outputs, combine AI models with deterministic processing and validation, and use measurement, regression testing and cost governance to improve reliability and operational efficiency. Experienced with Google Cloud AI Platform and modern AI coding environments including Codex, Claude Code and Antigravity IDE; comfortable taking solutions from problem definition through architecture, implementation, testing and deployment.

AI Platform Engineer / Founder at Image Prompt Maker (2025-01 – Present)

Architected, built and operated a production multi-model generative-AI platform spanning model orchestration, prompt engineering, data architecture, AI cost governance, product analytics and continuous deployment.

• Architected a provider-agnostic multi-model inference layer routing requests across 11+ image and language models, including Google Gemini, OpenAI image models, fal.ai-hosted models and Kling, through a unified interface.
• Designed model capability mapping, reference-image conditioning, asynchronous generation workflows, status polling and automated handling of long-running inference jobs.
• Built an LLM-driven 'Image DNA' pipeline using vision-model analysis to decompose uploaded images into 200+ structured attributes and reconstruct them into editable, deterministic generation instructions.
• Developed AI-assisted editing that combines LLM classification with deterministic transformation logic, JSON-schema validation, normalisation, deduplication and safeguards against unwanted attribute drift.
• Established AI evaluation and regression testing with a 220+ case automated test suite and live end-to-end smoke testing covering prompt behaviour, pricing accuracy and model-output consistency.
• Designed the PostgreSQL/Supabase data platform with Row-Level Security, version-controlled SQL migrations, database RPCs and structured data models; implemented 28 Deno/TypeScript serverless functions supporting inference, usage metering, billing, administration and analytics.
• Built multi-tier usage metering, global AI-spend controls and per-user rate limiting to govern unit economics and operational cost across paid AI APIs.
• Integrated PostHog funnel/cohort analytics, Stripe subscription billing and server-side conversion tracking to inform product and pricing decisions from behavioural data; operated deployments across Vercel, Cloudflare R2/CDN and cloud-hosted AI services.

AI & Data Automation Lead at Goftegoo (2026-01 – Present)

Architected and deployed the AI automation and data-processing layer of a multilingual news-intelligence platform, transforming high-volume unstructured social information into structured, deduplicated and decision-ready events.

• Built a cost-controlled LLM inference architecture with per-feature model routing, route-specific token budgets, prompt-cache-stable system instructions and hard daily AI-spend ceilings.
• Identified a reasoning-token configuration issue responsible for approximately 99% of completion-token expenditure on extraction workflows and redesigned routing to eliminate unnecessary reasoning usage while retaining an opt-in path for deeper inference.
• Designed a deterministic geo/time/keyword event-clustering and deduplication pipeline to identify reports relating to the same underlying event across independent information sources.
• Used production data and error analysis to tune classification behaviour; analysis of 181 events resulted in reclassification of approximately 55% of a previously miscategorised event class.
• Designed every AI call to fail safely to deterministic processing, preventing an LLM outage, timeout or malformed model response from breaking the user-facing data pipeline.
• Built the PostgreSQL/Supabase data platform with migration-controlled schemas, Row-Level Security and staged development-to-production promotion.
• Developed offline verification and regression gates using jsdom contract tests and headless Chromium browser validation; continuously evaluate where AI provides measurable value versus where deterministic software is more appropriate.

Freelance Translator at Freelance (2021-01 – 2023-12)

Dutch, Farsi and English translation; developed strong cross-language written communication and audience adaptation.

Sales Assistant & Cashier at C&A, Merksem (2017-01 – 2022-12)

Customer-facing retail roles requiring teamwork, reliability, communication and commercial awareness.

Sales Assistant at Tommy Hilfiger, Wijnegem (2023-01 – 2023-12)

Supported customers and day-to-day store operations in a commercial retail environment.

Bachelor Studies in Biomedical Sciences – University of Antwerp

Bachelor Studies in Pharmacy – Vrije Universiteit Brussel (VUB)

ASO - Wetenschappen-Wiskunde in Secondary education – Sint-Michielscollege