Field Notes
View full thesis document →EquationX explored the convergence of Cybersecurity, Artificial Intelligence, Machine Learning, and Network Security - with a focus on building systems that can intelligently detect, classify, and respond to threats in real time. Problem areas explored: - Intrusion Detection Systems (IDS) augmented with ML - moving beyond signature-based detection to anomaly-based detection that can identify novel attack patterns. - Adversarial robustness in security models - understanding how attackers can fool AI-based security systems and how to harden them. - Network traffic classification - using ML to distinguish benign from malicious traffic at scale, including encrypted traffic analysis. - Automated incident response - exploring how AI agents could triage and respond to security alerts, reducing analyst fatigue. Approach: The team combined literature review, dataset exploration (NSL-KDD, CICIDS), and rapid prototyping to test hypotheses about where ML adds the most value in the cybersecurity stack.
Notes & Reflections
“Do It Anyway.”
— Aarushi Singh
“Progress Over Perfection.”
— Akhila T
“Effort Never Dies.”
— Ashmitha Sri Anand
Domains
Margin Note
We started with more questions than answers, and somewhere along the way, learned that figuring things out together is half the fun.
Also in ABC 2026
Vigil
Cross-source signal correlation layer for manufacturing maintenance — correlates alarm logs, sensor trends, and operator observations in real-time to flag developing failures before they happen.
Influex
Creator intelligence platform bridging the gap between creators and brands — data-driven insights, authenticity assessment, portfolio management, and campaign performance tracking.
SpikeNest
Explored industrial sustainability and methane emission management, predictive maintenance using spiking neural networks, and attitude adjustment for CubeSats.
AgentOps
Explored AI-driven agentic solutions across Finance, Real Estate, EdTech, and MSME verticals — focused on identifying meaningful problems and validating real user pain points.
EdgeDaemon
Using AI to reconstruct, analyze, and improve decision-making processes — taking past decisions, reasoning, and outcomes to reconstruct how a decision was made and identify areas for improvement.
VoxNova
Explored AI reliability across Insurance & claims processing, Customer Service, Geospatial vision, and Warehouse logistics — mapping where AI breaks and how to make it trustworthy.
Horizon
Startup & problem-solving exploration across ML, Generative AI, and Medical domains — applying Mom Test principles, Reddit community research, and JEPA model exploration to discover real user pain points.