For a quarter century, IIIT Hyderabad’s annual R&D Showcase has been the campus’s open house for everything from autonomous cars to language technologies. This year, on its 25th anniversary, the institute turned the spotlight not on raw innovation, but on a more uncomfortable question: can we trust the systems we are building?
Under the theme “Trust in Technology – Security, Privacy and Transparency,” the showcase brought together over 400 demos across 29 centres and labs, yet the mood felt less like a science fair and more like a conversation about a social contract. Instead of celebrating clever hacks or benchmark-smashing models, teams were nudged to explain how their work behaves under stress, how it fails, and who gets hurt if it goes wrong.
Walking through the halls, you could sense a deliberate pivot away from the “move fast and break things” mentality that has haunted tech for over a decade. In the AI gallery, students demoed interpretable vision and language models where users could click through the features and training data that led to each prediction. A health AI group showed a triage model that not only flagged its top recommendation, but also surfaced a confidence score and alternative hypotheses—an explicit invitation for clinicians to disagree rather than simply defer.
Alongside them, a bias-auditing toolkit let visitors stress-test language models on caste, gender, and disability-themed prompts. Rather than hide uncomfortable outputs, teams presented them as starting points for design decisions: should we retrain, filter, or reframe the task? And who gets to decide those thresholds?
In the cyber-physical zone, model cars weaved through a mini smart city, overseen by a control stack built for graceful degradation instead of brittle optimisation. One demo showed how an autonomous system would behave when GPS was spoofed or when sensors “went blind” in heavy rain—switching to conservative modes, alerting remote operators, and logging events for post-hoc audit. A drone swarm setup illustrated secure coordination protocols, with visualisations of how one compromised node could be quarantined before infecting the rest.
Notably, security researchers shared space with formal methods experts, suggesting that IIIT-H sees “trust” as jointly about resisting adversaries and proving correctness in the first place. A live demonstration of model checking for an industrial controller, complete with a mock plant on the side, drew unusual crowds for something as arcane as temporal logic.
Privacy-preserving digital infrastructure emerged as the third pillar. One cluster of posters walked visitors through federated learning experiments with hospitals, in which patient data never left the premises while models still improved through shared gradients. Another illustrated differential privacy in plainer language than many policy documents manage: how a census-like system can answer aggregate questions about millions without exposing any individual’s record.
Researchers working on cryptographic voting, tamper-evident data logs, and transparent consent flows made the political implications explicit. They talked about technology as a space where citizens should be able to contest decisions—not just optimise preferences. Instead of glossy dashboards, their demos focused on the sometimes messy workflows needed to give people meaningful control over their data.
What tied these islands together was a quiet admission that metrics like accuracy, latency, and throughput—while still important—no longer suffice as primary measures of progress. Several faculty members noted that publishing a paper or shipping a product is now the midpoint, not the end, of the responsibility curve. Questions of who audits, who redresses harms, and how systems can be updated without destabilising societies were woven into presentations that, in previous years, might have confined themselves to the technical.
For the students, many of whom have grown up amid alternating waves of AI hype and AI panic, this framing seemed oddly natural. They are building careers in a world where facial recognition scandals, data breaches, and biased algorithms are front-page news. Their prototypes reflected an instinct that trust must be engineered from the outset rather than bolted on after a scandal.
For IIIT Hyderabad as an institution, the silver jubilee showcase read like a manifesto. It positioned the campus as a place that is willing to interrogate its own creations, where security, privacy, and transparency are treated as first-class research goals rather than compliance checkboxes. In a technology landscape increasingly shaped by regulation and public scepticism, that could prove as strategically valuable as any single breakthrough algorithm.
If this edition is a template, future R&D Showcases may be judged less by how futuristic the demos look and more by how honestly they face the limits and externalities of technology. Trust, IIIT-H seems to be arguing, is not a soft add-on to computing—it is the new hard problem.
-Raja Aditya
