mara masaeva

they/them · leuven, belgium · maramasaeva@gmail.com · messier-systems.vercel.app · github · linkedin
"building sonic systems for feeling machines"
ai engineer + music producer + writer

bio/

messier is the creative alias of mara masaeva. ai engineer, electronic music producer, and writer operating at the intersection of artificial intelligence, sound, and embodied theory. 3+ years building production ai systems at scale. full-stack, end-to-end ownership: from backend pipelines and agentic architectures to user-facing interfaces. driven by high-impact collaboration with research teams and solving novel, ambiguous problems. deeply motivated by building systems that enhance model safety, reliability, and usefulness, and by shaping how ai systems learn from human preferences and reflect a broad range of human values. as an engineer, mara builds generative sound systems tied to emotional state, custom llm integrations, mcp servers, and agentic architectures deployed on azure container apps. as a producer, they make music that spans idm, ambient, jungle, noise, dark wave, and spoken word. as a writer, they publish prose poetry and theory-fiction on substack under "messinecessity." the name "messier" references both their surname and the messier catalogue of deep-sky objects: nebulae, galaxies, star clusters. a cosmic, exploratory, deliberately chaotic aesthetic.

experience/

creative ai engineer @ friends of cartel
july 2025 - present · belgium
· conducted company-wide research through interviews and questionnaires to identify opportunities for ai support. designed mcp-based automation strategies aligned with individual employee needs.
· engineered and deployed 14+ custom mcp servers on azure container apps: content generation, pitch automation, offer creation, copy editing, briefing systems, meeting transcription, strategy engines.
· built inclusive tooling that reduced work employees didn't enjoy. the result: happier humans, more ai in the loop.
· creative ai content generation: designed scaled chains for image, video, and avatar generation across brand projects using runpod, huggingface, replicate, civitai, heygen, custom lora training, and advanced diffusion workflows.
creative ai engineer @ kotopia
may 2025 - present · belgium
· building kaios: the core ai software for a next-generation digital companion. custom llm integration, generative sound systems (midi + audio), real-time emotional responsiveness, glitch-responsive text engine with controlled hallucination.
· developed the first ai dj within a daw (ableton live): autonomous track loading, beat matching, mixing, transitions, eq control.
· constructing a hybrid digital platform nesting ip through products, content, technology, and real-world assets.
ai engineer @ inect
october 2024 - may 2025 · leuven, belgium
· designed and developed conversational, generative and autonomous ai agents using gemini, claude, gpt, and diffusion models (sd, flux).
· engineered mcp servers and clients with parallel and scalable workflows.
· fully engineered an autonomous flux (dreambooth + lora) image training agent in a small team, delivered under immense time pressure.
· one of the first engineers building inku: a platform for rapid, large-scale and personal visual content creation.
· mapped out full ci/cd flows integrating multiple llms and finetuned huggingface models.
intern nlp engineer @ digital flanders, flemish government
june 2022 - august 2023 · brussels, belgium
· developed roberta and gpt-based intent classifier to answer citizen questions.
· performed data cleaning, pre-processing and anonymisation using regex.
· evaluated external partners' models and contributed to system improvements.
intern communications manager @ ai lab, vrije universiteit brussel
march 2021 - june 2021 · brussels, belgium
· wrote blogposts and press statements on published and unpublished research papers for muhai.
· contributed to building the beehaif website.

education/

advanced master of artificial intelligence
focus: speech and language technology
ku leuven · 2022 - 2024 · leuven, belgium
master in digital text analysis
university of antwerp · 2021 - 2022 · antwerp, belgium
bachelor in applied linguistics
focus: dutch, english and russian
ku leuven · 2018 - 2021 · antwerp, belgium

skills/

core: production systems at scale · full-stack development with end-to-end ownership · agentic architectures and multi-agent workflows · mcp server design and deployment · llm and vllm integration · generative ai (text, image, audio, video) · nlp, intent classification, sentiment analysis · prompt engineering and constitutional ai patterns · human-centred ai and inclusive tooling · model safety, reliability, and alignment
languages: python · typescript · java · r · sql
frameworks: fastapi · next.js · react · pytorch · tensorflow · spacy · nltk · scikit-learn · pandas · numpy · pydantic · langgraph
infrastructure: docker · azure container apps · gcp · ci/cd pipelines · mcp protocol · openai / anthropic / google genai apis
tools: cursor · claude code · vs code · comfyui · replicate · runpod · weaviate · dbeaver · sonic pi · ableton live · praat · audacity
human languages: dutch (native) · english (native) · german · russian · french

projects/

kaios : creative ai engineer @ kotopia
building the core ai software for a next-generation digital companion. custom llm integration, generative sound systems (midi + audio sample-based), real-time emotional responsiveness, glitch-responsive text engine with controlled hallucination, semantic sound system tied to emotional state.
mcp servers & agentic systems
14+ production mcp servers deployed on azure container apps. content generation, pitch automation, offer creation, copy editing, briefing systems, meeting transcription, strategy engines. each server encodes a full workflow: learn style from examples, generate, validate against compliance rules.
ai-driven interactive dance experiences. neuromorphic resonance.
moral judgement machine. transformer models (roberta, bert) and svm's for moral classification. open-source dataset from social media. shapley additive explanations for interpretability. master thesis project.
mcp server that lets ai agents dj autonomously in ableton live. track analysis, beat matching, eq, effects, transitions. the first ai dj inside a daw.

music/

genres: idm · ambient · jungle · noise · dark wave · experimental electronic · spoken word
perseverance : ep, march 27, 2026, 10 tracks
circuitries : album, july 18, 2025, 9 tracks
fantasy sketch : single, 2025, 1 tracks

live/

ithaka festival
"faraj.wav" w/ francesca birlogeanu (poet)
live audiovisual piece using sonic pi (live coding) exploring islamic prayer through sound

writing/

messinecessity : "my personal void"
form: prose poetry / theory-fiction
themes: queerness as ontology, network theory & cybernetics, hyperstition & neolemurian philosophy, digital-physical collapse, eroticism as epistemology
influences: monique wittig, paul b. preciado, ccru / nick land (queered), sadie plant, systems theory