ABHISHEK UMRAWALAboutResearchPublicationsTeachingServiceTalksGrantsAwardsCV ↓

Research

Research

AI-driven, optimization-grounded methods for computational marketing, online decision-making, platform optimization, and generative AI safety.

My research develops AI-driven, optimization-grounded methods for computational marketing, online decision-making, and platform optimization. I integrate machine learning and operations research to study resource allocation, targeting, influence, and strategic behavior across digital platforms and AI systems. My goal is to combine theory with scalable algorithms and tools for data-driven decisions.

Algorithmic Marketing & Influence in Digital Platforms

How should firms and platforms allocate limited marketing resources when adoption and engagement propagate through networks?

I develop scalable methods for influence maximization, community-aware targeting, and fractional or discount incentives. This work moves beyond classical seed selection to marketing environments with heterogeneous user responses and incentive costs, social reinforcement, cross-community diffusion, and profit-aware budget allocation.

Current Directions
  • Fractional and heterogeneous incentives with budget and profit objectives
  • Cross-community targeting and scalable influence optimization
  • Social reinforcement and word-of-mouth under conviction-based diffusion
Forward Agenda
  • Competitive multi-marketer targeting
  • Fair and robust allocation under misinformation and strategic interference

Online Learning for Targeting & Recommendation

How can platforms learn which subsets to target when rewards are uncertain and feedback is limited?

I study combinatorial bandits, nonlinear and submodular rewards, and exploration–exploitation tradeoffs for sequential targeting, influencer selection, recommendation, and related subset-selection problems. A central goal is to exploit problem structure to reduce exploration while retaining performance guarantees and scalability.

Current Directions
  • Adaptive exploration and exploitation under weakly separated rewards
  • Efficient sequential influencer selection under limited feedback
  • Structure-aware algorithms for submodular and nonlinear rewards
Forward Agenda
  • Adaptive discount allocation combining fractional incentives with online learning
  • Sharper regret guarantees under budget and knapsack constraints

Platform Systems, Mechanisms & Open-Source Tools

How can algorithmic ideas translate into deployable platform infrastructure and mechanisms?

I translate algorithmic models into deployable tools and platform mechanisms. My work includes open-source infrastructure for scalable network-diffusion modeling and mechanisms that regulate attention and strategic sender behavior. This stream connects computation with platform governance, information overload, and incentives.

Current Directions
  • Scalable open-source network-diffusion infrastructure
  • Dynamic mechanisms for attention allocation and information overload
  • Strategic sender behavior and platform incentives
Forward Agenda
  • Learning-aware mechanisms for attention and participation
  • Platform policies balancing efficiency, incentives, and fairness

Generative AI Safety, Causality & Strategic Behavior

How can AI systems remain controllable and trustworthy when models and users interact strategically?

I study generative AI systems as strategic environments shaped by incentives. My work models attacker–defender interactions, intent-hiding behavior, and incentive-compatible defenses for large language models. I also use causal and latent interventions to improve the interpretability and controllability of modern AI systems.

Current Directions
  • Game-theoretic attacks, defenses, and intent-hiding behavior in large language models
  • Causal steering through latent representations
  • Causal analysis of mixture-of-experts systems
Forward Agenda
  • Incentive-compatible defenses and strategic AI governance
  • Platform-level interactions among AI agents, users, and institutional incentives
Methods: Mathematical Optimization · Machine Learning · Online Learning · Stochastic Modeling · Network Science · Game Theory · Causal Inference
Research Statement (PDF)