I build computational models of how audiences watch films.
I study how people perceive, interpret and remember cinematic media, combining perception models, narrative structure analysis and interactive systems to understand how films guide attention, emotion and interpretation.
Attention I treat as a perceptual mechanism and trust as a higher level interpretive judgement, and I study how computational media systems shape the relationship between the two.
Computational media researcherLondon
Artificial Viewera multimodal project with Professor Tim SmithCreative Computing Institute, University of the Arts London
loading
Recent
What is happening
2026
A paper under review with SCSMI, Narrative DNA, on structural patterns in film scripts.
2026
Invited lecture at the Creative Computing Institute, UAL: The Mathematics of Cinema.
2026
Presented DinoSynthesis at the UAL and CHEAD Generative AI Event, London.
2026
Runner up, Makerversity Under 30s, Somerset House. Second place, UAL CES and Lovable Showcase Gala.
2025
Began the MRes in Creative Computing at UAL on a competitive scholarship of twenty five thousand pounds.
Research
Three strands
The long term goal is artificial systems that can watch a film, find the
narrative structure in it, and model viewer engagement the way a human
audience does.
Computational film perception
How visual attention, emotion and narrative structure shape viewer
engagement, studied through behavioural signals, saliency modelling
and temporal media analytics.
Narrative structure intelligence
Models that pull structural patterns out of film scripts and
sequences, phases, transitions, tension, causal density, and relate
narrative form to how an audience receives it.
Generative and interactive media
Experimental systems for storytelling, performance and AI mediated
media, including computational treatments of traditional performance
forms and questions of trust in autonomous media.
Systems
Built as research instruments
Built as research instruments rather than as products.
2025 to 2026
SPECT
Computational spectatorship modelling
The umbrella framework for the spectatorship research: modelling viewer perception, attention and narrative understanding in film as one system rather than as separate measurements. It combines saliency modelling with multimodal perception signals and behavioural experimentation, and gives the individual studies a shared structure to report into.
Computational modelling of audience attention across a runtime
Narrative engagement and interpretation modelling
Saliency systems driven by multimodal perception signals
Behavioural experimentation as empirical grounding
Most saliency work flattens a film to a 2D heatmap, which discards the dimension cinema is composed in: depth. This pipeline augments saliency with depth estimation to extract attention as a trajectory through 3D space, then clusters those trajectories to find where perception groups and where it scatters across a sequence.
Depth augmented saliency extraction
3D attention trajectory reconstruction across sequences
Borrows its method from sequence analysis in biology. Film scripts are encoded as symbolic sequences, then mined for recurring phases, transitions and structural motifs, the way a genome is read for genes and regulatory patterns. The output is a structural fingerprint of a film that can be compared across a corpus and correlated against audience reception. Basis of the SCSMI 2026 submission.
Symbolic encoding of screenplay structure into analysable sequences
Narrative phase extraction and sub genre transition detection
Structural motif mining across a film corpus
Sequence analysis, NLP, clustering. Python.
2025 to 2026
Gradient.ai
Autonomous news intelligence and trust
A multi agent news system designed to surface bias, uncertainty and perspective by presenting several algorithmic framings of the same story. It is both a working pipeline, aggregating, cross checking, validating and generating the broadcast, and a research instrument for studying how people interpret and trust news that a machine assembled.
Multi agent aggregation and cross checking pipeline
Transparency oriented validation, with several framings side by side
Behavioural user studies on trust and interpretation
Multi agent systems, LLMs, RAG, behavioural experimentation. Python, Next.js.
2025 to 2026
ShadowStage
Computational Tholpavakkoothu
Tholpavakkoothu is a centuries old Kerala shadow puppetry ritual. I treat that cultural form not as inspiration but as a computational system that already exists: one with rules, constraints and grammars encoded in its practice. ShadowStage formalises those grammars into a programmable motion system, adds semantic light control, and puts a human performer and a model in the same loop rather than replacing one with the other.
Joint based motion grammars derived from traditional practice
Semantic lighting control tied to dramaturgical state
A human and AI co performance loop that preserves the ritual form
A network for researchers. It surfaces people by the work they do,
so that opportunity reaches talent wherever it happens to be, rather
than only where someone already knows to look.
MRes Creative Computing, Creative Computing Institute, UAL
On a competitive scholarship of twenty five thousand pounds. Computational media, cinematic perception, narrative intelligence, multimodal AI.
2022 to 2024
MTech Computer Science and Engineering, AI, Digital University Kerala
Outstanding Student Award for the batch.
2021 to 2022
PG Diploma, Data Science and Analytics, Kannur University
2019 to 2021
MSc Physics, Kannur University
Experience
Where the research has been
Dec 2024 to Sept 2025
AI Research and Development Engineer, Digital University Kerala
Multimodal AI systems across language models, computer vision and AI assisted verification. LLM pipelines, multi agent workflows and retrieval augmented architectures.
Jul 2024 to Sept 2025
Co founder, KnowLumi
An education platform built around adaptive interview simulation. Designed and deployed a multi agent AI interviewer.
Jul to Dec 2023
Research Intern, IIM Kozhikode
Language models and enterprise information systems.
Apr to Jun 2021
Research Intern, CSIR National Physical Laboratory, New Delhi
Python, R, React, Next.js, Flask, Git, SQL, API integration, workflow automation, Linux
Visualisation and research tools
Matplotlib, Seaborn, Plotly, Tableau, Power BI, Jupyter, Google Colab, Excel
Awards
Recognition
UAL scholarship of twenty five thousand pounds for the MRes, 2025
Outstanding Student Award, Digital University Kerala
Runner up, Makerversity Under 30s, Somerset House, London, 2026
Second place, UAL CES and Lovable Portfolio Showcase Gala
First prize, Smart India Hackathon 2023, university level
First prize, Sharktank Hackathon, Google DSC WoW Kerala
Vidyadhan Scholarship, funded by Infosys co founder S D Shibulal
Contact
Say hello
I am glad to talk about computational perception, narrative structure, AI mediated media, or collaborations between them. Open to research collaboration and PhD conversations.
