November 20, 2026

UK News Latest: A Need to See IT Publishing Site

Sharing Business, Lifestyle and Community News

closeup photo of eyeglasses

DreamWare Hackathon 2025: Enterprise Engineers Evaluate Emotional AI Applications Built in 72 Hours

Industry veterans from Microsoft, AWS, Okta, and AliExpress assess whether creative AI applications can meet production security, infrastructure, and architecture standards


Hackathon Raptors, a UK Community Interest Company (CIC #15557917), concluded its DreamWare Hackathon on December 1st, 2025, with results highlighting the gap between creative AI ambition and production-ready implementation. The 72-hour competition challenged 29 teams to build applications where “interfaces behave organically, systems remember feelings, and digital experiences feel dream-like”—a theme the organization termed “Engineering the Surreal.”

The evaluation framework weighted Technical Execution at 40%, Concept & Depth at 30%, and Originality & Presence at 30%. This structure prioritized whether creative visions could survive contact with production requirements: security, scalability, reliability, and architectural coherence.

Expert Panel Applies Enterprise Standards to Creative Applications

The judging panel brought expertise spanning identity systems, cloud infrastructure, agentic AI, mobile architecture, MLOps, government systems, and fintech development. Their evaluations assessed not just what hackathon projects demonstrated, but what production deployment would require.

Arun Kumar Elengovan, Senior Security Engineer at Okta with 13+ years protecting enterprise identity systems, evaluated submissions through a security lens. His experience building threat detection systems and implementing secrets management informed assessment of how emotional AI applications handle sensitive user data.

“Emotional data is arguably more sensitive than financial data,” Elengovan observed. “If someone steals your credit card number, you get a new card. If someone accesses your emotional history—your anxieties, your struggles—that exposure is permanent. Most submissions applied startup-grade security to government-grade sensitive information.”

Projects like “The Garden of Forgotten Feelings,” which stores emotional data in browser localStorage, illustrated the pattern: technically functional for demonstration, but lacking encryption, backup, or access controls appropriate for production deployment with real user data.

Veera V S B Nunna, Senior Data Engineer at AWS and principal technical lead for Knowledge Engineering, brought expertise in RDF/OWL ontologies and knowledge graphs powering LLM applications. His Amazon Inventor Award for ontology-based service dependency modeling informed evaluation of how creative applications structure emotional understanding.

“Traditional databases store data points. Knowledge graphs capture relationships,” he explained. “When an AI companion needs to understand that a user’s current anxiety connects to work stress mentioned three weeks ago, that requires semantic modeling most hackathon projects don’t implement. They store text; they don’t model meaning.”

His assessment framework examined whether projects like “ECHOES”—an AI-powered emotional sanctuary—implemented genuine emotional memory through knowledge architecture or simulated understanding through prompt engineering.

Giridhar Raj Singh Chowhan, Principal Architect for AI Agent Platform at Microsoft with 14 years building autonomous systems for Dynamics 365, evaluated whether submissions demonstrated genuine agentic architecture or sophisticated prompt wrappers.

“The defining characteristic of an agent isn’t intelligence—it’s autonomy in pursuit of goals,” Chowhan noted. “A chatbot waits for input, generates output, and stops. An agent observes continuously, maintains state across interactions, forms plans, and executes actions. Most emotional AI applications are chatbots claiming to be companions.”

His work on Store Operations AI Agent and Personalized Shopping AI Agent—systems that interpret ambiguous situations and take coordinated action—provided context for evaluating whether creative applications implemented perception layers, memory systems, and planning mechanisms rather than request-response patterns.

Egor Korotkii, iOS Developer at AliExpress who created Fusion, a Backend-Driven UI engine enabling app releases without App Store deployment, assessed whether creative applications could evolve without deployment bottlenecks.

“Applications promising dream-like experiences that shift based on user emotion face a fundamental problem: how do you build software that feels alive when every change requires week-long approval cycles?” he asked. “Server-driven UI architecture solves this—the interface becomes data, streamed and rendered dynamically. Most submissions hardcoded every emotional-visual mapping, creating exponential complexity.”

His experience at VK (leading feed refactoring for the main application) and Mail.ru Group informed evaluation of whether projects could maintain the fluid, evolving interfaces their concepts promised.

Vladyslav Haina, Lead MLOps & Cloud Architect specializing in real-time AI infrastructure, evaluated whether emotional AI applications implemented the streaming architecture their promises require.

“Most teams built request-response systems,” Haina observed. “User sends input, system processes, system responds. That architecture cannot support real-time emotional awareness. The system only knows what you feel when you explicitly tell it. It cannot observe continuously or detect distress from typing patterns.”

His expertise with Kafka, Flink, and vector databases (Pinecone, Weaviate, Milvus) informed assessment of projects like “The Living Dreamspace,” which promises music reactive to emotional states inferred from typing patterns—functionality requiring event streaming infrastructure most hackathon projects lack.

Oleksandr Sosnytskyi, Senior Software Engineer at Knot (New York fintech) with four years mentoring developers at Hillel IT School and 40+ production projects as Technical Lead at Wideweb, evaluated submissions through the lens of tutorial-to-production gaps.

“The junior developer’s code works perfectly—on their laptop, with test data, under ideal conditions,” he reflected. “Then it meets production: real users, unexpected inputs, traffic spikes, edge cases the tutorial never mentioned. The elegant solution becomes a debugging nightmare.”

His teaching experience informed recognition of which architectural decisions would scale and which contained hidden fragility—the difference between promising prototypes and deployable products.

Damodhara Palavali, Lead Developer at the Social Security Administration handling benefits data for 70+ million Americans, brought government-grade data protection standards to evaluating emotional AI applications.

“In government systems, you assume adversaries are sophisticated and persistent. Nation-state actors target citizen data. The threat model isn’t a curious teenager—it’s organized, well-funded attacks,” he explained. “Emotional AI applications collecting vulnerability data should apply similar threat models. Most don’t.”

His IEEE Senior Member status and recent presentations at IEEE ICCST 2025 (International Carnahan Conference on Security Technology) informed evaluation of whether projects implemented the audit trails, access controls, and encryption that federal systems require—standards most emotional AI applications ignore despite handling comparably sensitive data.

Technical Patterns Across Submissions

The 29 submissions revealed consistent patterns in how creative applications approach—or avoid—production requirements.

Security Architecture: Most projects treated security as optional rather than foundational. Emotional data stored without encryption, API keys embedded in client-side code, and third-party AI integrations without data sanitization appeared across multiple submissions. Projects making wellness claims face particular scrutiny: applications promising therapeutic value must implement crisis detection, escalation protocols, and appropriate boundaries.

Real-Time Infrastructure: Projects promising responsive, adaptive experiences typically implemented request-response patterns incapable of continuous observation. Event streaming architectures (Kafka, Flink) that enable genuine real-time emotional awareness remained rare. The gap between “reacts to emotions” and “observes emotions continuously” represents significant infrastructure investment most teams didn’t anticipate.

Memory and Context: Emotional AI companions promising to “remember” users typically implemented session-scoped context or localStorage persistence rather than semantic knowledge structures. The difference matters: genuine emotional memory requires connecting current states to historical patterns through meaning, not keyword matching.

Agentic Behavior: Applications described as AI companions typically demonstrated chatbot architecture—waiting for input rather than observing, responding rather than anticipating, generating text rather than taking action. Genuine agent architecture requires perception layers, memory systems, planning mechanisms, and action interfaces.

Evaluation Criteria and Methodology

The weighted evaluation framework reflected production deployment priorities:

Technical Execution (40%) assessed implementation quality, architectural decisions, security practices, and scalability potential. Judges evaluated whether code organization, error handling, and resource management demonstrated production discipline or hackathon expedience.

Concept & Depth (30%) examined whether creative visions had coherent underlying architecture. A concept promising emotional memory achieved depth only if underlying data structures actually supported persistent, evolving understanding rather than session-scoped prompt injection.

Originality & Presence (30%) evaluated whether applications felt genuinely novel and whether systems demonstrated personality through architecture rather than prompt engineering. Systems that observe, anticipate, and proactively engage feel present; systems that only respond feel passive.

Industry Context and Future Implications

The commercial landscape for emotionally intelligent applications is expanding rapidly. AI companions, wellness chatbots, therapeutic support systems, and creative assistants attract substantial investment and user adoption. DreamWare submissions represent early attempts at applications that major technology companies are actively developing.

The gap between hackathon demonstration and production deployment is where most creative AI projects fail. The panel’s enterprise experience—spanning identity systems protecting millions of users, cloud infrastructure serving global applications, and government systems meeting federal compliance requirements—provided perspective on what production deployment actually requires.

“The creative visions at DreamWare were genuinely impressive,” noted Palavali. “Teams imagined applications that understand human feelings in ways that could genuinely help people. Making those visions trustworthy requires treating emotional data with the seriousness it deserves.”

The teams that successfully commercialize these visions will be those that pair creative ambition with engineering discipline—building security, infrastructure, and architecture as carefully as they design user experiences.

About Hackathon Raptors

Hackathon Raptors operates as a UK Community Interest Company dedicated to organizing technology challenges demonstrating practical applications of software engineering methodologies. Based in London and registered as CIC #15557917, the organization has conducted multiple successful hackathons attracting participants from 30+ countries.

The organization’s fellowship model brings together engineers from Google, Microsoft, Amazon, Meta, and other technology companies, requiring peer recognition for expertise in their fields. This approach ensures evaluation by practitioners who understand production system requirements rather than theoretical ideals.

DreamWare Hackathon 2025 featured $2,300 in prizes distributed across winners and category awards. More information is available at raptors.dev or dreamwarehack.com.