Ayush Manish Agrawal. Curious first. Then I build it.
I work on production multi-agent systems for cloud infrastructure operations at Microsoft. The best way to see what I like to build is to meet one: my agent is right below. Ask it anything about me and watch it dig up the receipts. Sound on is more fun.
Screen Ayush.
About 6 minutes · no score · you keep the notes
Agenda · Background · Experience · Projects · Fit check · Anything else · follow it or skipYour pick is remembered. Switch mid-session with one control.
Architected and built CIS Copilot (ARLI), a multi-agent system for asking about buildout status and incident impact in plain language. Now owns the architecture for CIS incident automation, and collaborates on Microsoft Scout.
Details
Owns the high-level architecture, integration and orchestration for CIS incident automation: triage, categorization and mitigation workflows that reduce manual incident handling effort by up to 60%, improve SLA compliance, and support faster resolution of high-severity incidents.
Architected and built CIS Copilot (ARLI), an enterprise-scale multi-agent system that lets planners, SREs and leadership query real-time buildout status, incident impact and post-analysis insights across CIS platform data in natural language.
Designed domain-specialized LLM tooling for the Copilot: schema-aware indexing, few-shot retrieval, fuzzy entity matching, and business-rule driven query enforcement.
Implemented autonomous NL2SQL and real-time visualization workflows: agents generate SQL, execute queries and render Mermaid-based operational visualizations, reducing reliance on manually maintained dashboards.
Delivered ML-driven ETA prediction for GPU buildouts: 79% accuracy within ±7 days, outperforming SLA-based estimates in 83% of cases, with incident-aware rolling updates at sub-hour latency.
Collaborating cross-team on new features for Microsoft Scout, Microsoft's always-on enterprise AI agent, to expand its autonomous, agentic capabilities.
Software Engineer · University of Nebraska – Board of Regents
A Python library that automated versioning and tagging on merge requests, and a full-stack .NET and React issue tracker.
Details
Built a Python library automating project versioning and tagging on incoming merge requests, improving the team's DevOps lifecycle by 30%.
Architected and delivered a full-stack issue-tracking application (.NET Core, React) with hardened input validation, allow-listing and reCAPTCHA, which drove a 55% increase in user issue reporting.
2020 — 2022
Research Engineer · Manifold Computing · OpenMined
Open research on how deep networks learn and on federated learning. Contributed talks at NeurIPS 2020 workshops.
Details
Manifold Computing (Sep 2020 – Sep 2022): designed a relative weight change (RWC) metric to interpret learning dynamics in deep networks using established computer vision architectures as baselines, ran the experiments in PyTorch, and presented at the NewInML workshop at NeurIPS 2020 (contributed talk).
OpenMined (Jul 2020 – Dec 2020): studied ML systems from a data-privacy perspective, reproduced federated learning algorithms in PyTorch, and proposed a metric for the complexity of federated learning systems (contributed talk).
Research
03 — Papers
Accepted · Microsoft MLADS 2026 · MSJAR 2026
Grounding Enterprise Text-to-SQL in Tribal Knowledge
Introduces table rules, retrievable per-table constraint files written by domain experts, as a complement to schema and few-shot retrieval in NL2SQL agents.
Hiring for applied AI or agent systems? Reach out with the role and I’ll reply myself. I’m also open to collaborations on agent systems, side projects that need a builder, research, and the occasional talk. I read everything.
