AI · difficult documents · cultural heritage · photographs
I study how machines read imperfect traces.
I work at the intersection of NLP, document analysis, digital humanities, and cultural heritage. I build and question AI systems that try to read noisy, visual, multilingual, historical, and politically situated documents.
About
Mostly serious work. Not always serious tone.
I do research on machines that read difficult documents: historical newspapers, OCR and HTR transcripts, handwritten archives, epigraphic inscriptions, administrative records, financial documents, multilingual media, and multimodal cultural heritage collections. I am interested in what AI can extract from them — people, places, events, relations, narratives, uncertainty — and also in what it misses, invents, or quietly damages.
Current obsessions
Small things currently taking too much space in my brain
OCR ghosts
Errors, distortions, damaged text, and the strange poetry of machine mistakes.
Follow the ghosts → NO. 2Temporal entities
Names, places, roles, aliases, and meanings that change across decades or centuries.
Open time → NO. 3LLMs in archives
Fluent answers, weak evidence, historical hallucinations, and models pretending to understand time.
Read notes → NO. 4Visual evidence
Photographs, layout, bodies, rooms, objects, traces, and personal space as a stubborn archive.
See photo projects →Selected work
Things I keep returning to
A compressed map of recurring research themes. Each box opens a project page with the work behind it and selected papers.
Noisy Documents
I study OCR, HTR, digitization errors, degraded text, and how noise changes downstream meaning.
Research + papers → 02Semantic Extraction
I work on named entities, entity linking, relations, events, and the small structures that make documents searchable.
Research + papers → 03Historical AI
I care about temporal modelling, multilingual archives, and language that refuses to stay stable.
Research + papers → 04Cultural Heritage
I work with digital epigraphy, Armenian and Ukrainian heritage, cultural weaponization, and structured memory.
Research + papers → 05Multimodal Archives
I connect text, image, layout, typography, photographs, and visual traces in document collections.
Research + papers → 06Applied Document Intelligence
I have also worked on document fraud, fake news detection, epidemic monitoring, and large-scale research infrastructures.
Research + papers →Photographs / video
Photo projects
Extracted from my art projects since 2009: project titles, dates, media, descriptions, captions, and plates. Click a cover to open the full project.
01
Space in Time
Private space, performance, rooms, projection, exhibition.
2010 · performance · video · photography · programming · interactivity
02
2007 Journal
A book of images, text fragments, truthfulness and staring.
2007 · photography · book
03
Christmas '10
Small events, familiar surroundings, memory loops.
2010 · photography
04
Tableau Vivant
Photography and exhibition views, 2009.
2009 · photography
05
Tătărași
Photography and exhibition views, 2009.
2009 · photography
06
A reface
Video stills from the 2009 portfolio.
2009 · video
07
The same
Video project, 2009.
2009 · video
08
Înainte de a adormi
Video project, 2010; “cu 30 de ani înainte.”
2010 · videoWriting
Posts, notes, opinions
A place for short texts about AI, archives, cultural heritage, evidence, noise, and other things that refuse to stay neatly technical.
LLMs in archives: fluent, useful, suspicious
On why good prose is not the same thing as grounded historical understanding.
NoteDocuments are evidence, not just data
A small manifesto for building AI systems that keep traces attached to claims.
Contact