Can a conversation
be mistaken for you?
A two-phase study of suspicion, trust, and identity detection in conversational AI, using user-trained clones as a way to investigate human–agent interaction.
Explore the studyHuman–AI interaction · Product · Operations
Make complex
systems feel
human.
I’m Shriya, an HCI researcher and product practitioner working across the places where people, AI, and real-world systems meet.
Signals in conversation / 01One practice,
several lenses.
I move between observing how people make sense of technology, shaping products around what they need, and building the operational systems that help teams use AI responsibly. The through-line is the same: make the work legible, test assumptions with people, and carry what we learn into the next decision.
01 / Selected work
Questions made
into practice.
Different settings, connected by a habit of turning ambiguity into something a team can inspect, test, and improve.
Can a conversation
Making port release
workflows clearer.
Product work across a German Ports release-order flow and related operational features, from training and slot changes to dashboards and incident handling.
View contribution areasFrom scattered knowledge
to governed AI use.
Building the foundations for useful, reviewable AI support: knowledge audits, modular content, team enablement, and a clear path for skills to be reviewed before sharing.
View contribution areasWhat if a survey
read like a story?
Exploring “comicification”: redesigning surveys as sequential visual narratives, then bringing the format into workshops and public-facing research experiences.
See the project02 / Ongoing research
Cloned Testing
A thesis project on suspicion, trust, and identity in conversations with people and AI agents.
M.Sc. Human–Computer Interaction · Bauhaus-Universität Weimar
When a familiar voice
isn’t the person.
Cloned Testing is a hybrid method for evaluating human–agent interaction using knowledge-garden-based user clones. The study asks how people decide who—or what—they are talking to, how suspicion changes as a conversation unfolds, and what signals shape that judgment.
The work is structured as two linked activities: people first train a conversational clone through persona role-play, then evaluate conversations with a human operator, a baseline AI, and a clone trained from their material. A mid-session switch makes identity judgments part of the interaction rather than a question asked only at the end.
Evaluation combines final identity guesses and confidence with time to decision, suspicion changes, switch detection, conversation traces, accuracy, and repair behavior. The aim is a method that makes those judgments observable and useful to designers of conversational systems.
03 / In the field
Work that meets
the everyday.
Product and operations are where a good idea has to survive actual constraints, handoffs, and human habits.
Product · Passify
Operational product work
Contributed to product initiatives for German port logistics, working across workflows used by companies and operational teams.
- German Ports release-order workflow
- Training module and slot split / merge
- Dashboards and incident management
- VAT campaign supporting 220–250 companies
AI support · voize
AI people can work with
Helped structure how internal knowledge and AI support fit together, with attention to quality and clear ownership.
- Intercom retention KPI work
- Notion knowledge-base audit and modular structure
- Claude enablement for teams
- Department review before AI skills are shared
Research through design · CSL
Surveys as visual stories
Co-founded Comic Survey Lab with Konstantina Marra to explore how sequential visuals can change the experience of answering research questions.
- Four workshops and a 2024 pilot
- Two 12-scene story decks and nine technique cards
- Installation at the New European Bauhaus Festival
- Mensch und Computer 2025 poster
04 / Experience & formation
Across teams
and disciplines.
Computer science gave me a systems foundation; HCI gave me methods for understanding the people inside those systems.
voize GmbH
AI Support & Knowledge Manager
AI support, knowledge systems, and internal enablement in a health-tech startup.
Passify GmbH
Werkstudent, Product Management
Product work for port-logistics workflows and operational tools, including release orders, training, slot handling, dashboards, and incident management.
NetBluez
Product & design experience
Selected work and project details to be added.
Bauhaus-Universität Weimar
M.Sc. Human–Computer Interaction
Research methods, ubiquitous computing, usability engineering, physiological computing, visualization, and comics in East Asia.
Rajasthan Technical University
B.Tech, Computer Science & Engineering
Foundation in software engineering, databases, algorithms, AI/ML, graphics, and digital image processing.
05 / How I work
Make the invisible
discussable.
Start with a real question
Turn a broad concern into a decision, behavior, or experience that can be observed.
Make the system legible
Map the people, information, and handoffs so everyone can see where friction lives.
Close the learning loop
Carry what research reveals into product choices, team processes, and the next test.