AI Systems Engineer | Healthcare AI Researcher | Agentic AI Builder

Welcome to my GitHub. I build AI systems that operate under real-world constraints: privacy, distributed data, limited compute, and deployment complexity.

I sit at the intersection of AI Research and Systems Engineering, focusing on:

• Healthcare AI: Generative models for medical time-series, federated learning for hospitals
• Agentic Systems: Multi-agent orchestration, workflow automation, tool-calling frameworks
• Neuromorphic Computing: Brain-inspired efficient computation for resource-constrained environments
• Cloud Infrastructure: AWS, containerization, production ML systems
• Sustainability Intelligence: AI for ESG compliance and enterprise carbon accounting

My goal: Translate research into production-grade systems.

Languages	AI/ML	Frameworks	Infrastructure

Specializations: Generative AI, Agentic Systems, Healthcare AI, Neuromorphic Computing, Sustainability Intelligence

Under Prof. Jitin Singla

Building age-conditioned diffusion models (SSSD-ECG) for ECG signal generation to address dataset imbalance across age groups.

• Dataset: PTB-XL (21,837 ECG recordings)
• Architecture: WaveNet residuals + S4 layers
• Impact: 15–20% improvement in classifier robustness
• Key Learning: Domain-aware evaluation matters more than loss functions

Under Prof. Ankush Kumar

Electrode selection framework for nanowire networks using information-theoretic measures (mutual information, transfer entropy, graph centrality).

• Hardware: ~1500 nanowires, ~2636 memristive junctions, 16 electrode grid
• Result: 5× hardware reduction (16 → 5 electrodes), performance retained
• Validation: Sequential MNIST, NARMA-10, Mackey-Glass prediction, XOR Delayed

📍 Key Insight: Efficient systems require understanding information flow, not just raw compute

Privacy-preserving collaborative learning across hospitals

• Scale: 5 hospital nodes, ~26,801 samples
• Modalities: ECG, clinical vitals, chest X-rays
• Architecture: Flower federated learning, differential privacy (ε≈5), blockchain audit trail
• Models: S4, TabNet, MLP, ResNet50, Transformer fusion
• Results: +25.3% AUROC (geriatric cohort), +14% AUROC (general ECG)

Key Learning: System design matters as much as model architecture.

Move beyond chatbots to operational AI systems

Currently building multi-agent orchestration for financial workflows:

• Agent types: Research, Market Data, Financial Analysis, Reporting, Execution
• Tech stack: FastAPI, MCP integrations, AWS infrastructure, tool calling
• Vision: Production-grade financial intelligence platform with real workflow execution

AI-native operating system for ESG compliance

Platform modules:

• Carbon accounting & emissions tracking
• ESG reporting automation
• Compliance monitoring
• Supplier sustainability intelligence
• Energy optimization predictions

Tech: Agentic AI, enterprise RAG, regulatory intelligence, document understanding

See below for organized research + production projects.

Founder & Lead, AWS Cloud Club

• Built community of 400+ students
• Organized technical workshops and AWS Cloud Quests
• Hosted learning challenges with AWS Community Builders
• Participated in AWS Summit Bengaluru
• Mission: Help students move from AI experimentation to deployable systems

• Technology initiatives for social impact
• Volunteer management systems
• STEM awareness workshops (government schools)

• Technical workshops and mentorship
• Project guidance and community building

• Oracle Cloud Infrastructure Generative AI Professional (2025)
• Oracle AI Cloud Database Services Professional
• Harvard CS50 Python
• Cisco Programming Essentials

1. FINCHAT: Building production multi-agent systems for financial workflows
2. Healthcare AI: Scaling medical time-series research to real hospital deployments
3. AWS & Cloud: Deepening cloud infrastructure expertise for production ML systems

• Email: aishwaryajacse@gmail.com
• LinkedIn: [https://www.linkedin.com/in/aishwarya-j-a-b2b5a133a/]
• Research Interests: Always open to discussions on healthcare AI, agentic systems, and neuromorphic computing

Last updated: June 2026