Krishna Chaitanya Balusu

Software Engineer · Meta Platforms

Software engineer with 8+ years of experience across Meta, Microsoft, and financial services. Currently building AI-powered autonomous incident response systems at Meta. Open-source maintainer and speaker on AI agent reliability, with a long-running focus on accessibility and applying distributed-systems principles to AI infrastructure.

krishnabkc15@gmail.com github.com/Krishnachaitanyakc linkedin.com/in/kcbalusu

Google Scholar ORCID OpenReview DBLP

Experience

Meta Platforms

July 2022 – Present

Software Engineer

OpsMate — Autonomous Incident Investigation Sept 2025 – Present

Meta's autonomous incident investigation agent, which the program reports investigates roughly 1,500 incidents per day.

  • Designed and shipped the evaluation infrastructure — the skill-attribution evaluation with its automatic grading and logging pipeline, plus extended structured capture of human root causes — making agent output systematically gradable across the program. Listed author on “Evaluating OpsMate: Grading AI Incident Investigations at Scale,” @Scale AI & Data 2026.
  • Built the mechanisms by which the agent recovers from its own failures without a human in the loop: the improve workflow and its batch mode, the determinism and safety hardening that followed, and a generalized inconclusive-retrigger promoted into the shared adaptive-investigator trait, so every investigation framework inherits automatic reinvestigation rather than getting it one surface at a time.
  • Built the observability stack, including a lifecycle investigation skill that lets any engineer trace a failed agent run to the step that failed — silent failure being the characteristic failure mode of LLM agent systems, where the system reports success while producing nothing.

Instagram Feed & Reels Dec 2023 – Sept 2025

Spearheaded latency and reliability optimizations for Instagram Feed and Reels, driving improvements in user engagement and platform performance at scale.

Data Security Systems July 2022 – Dec 2023

Improved UX for a critical data-erasure and destruction system used across Meta's data centers, making technicians more efficient while maintaining the highest data-security standards.

Microsoft

July 2018 – May 2022

Software Engineer

  • Led design and implementation of cross-organization communication in Microsoft Teams.
  • Integrated Teams with Azure Blob Store for caching, achieving <100 ms response times.
  • Optimized user-membership queries by 80× (40 s to <500 ms).
  • Designed micro-services for user metadata storage.
  • Implemented accessibility features and fixes in Microsoft PowerApps, and authored a published whitepaper on PowerApps accessibility guidelines.

Edelweiss Financial Services

July 2017 – June 2018

Machine Learning Intern

  • Built ML models for market-volatility analysis and risk assessment.
  • Developed a loan-application evaluation model combining personal financial data with macroeconomic indicators.
  • Forecast India's inflation rate with neural networks, comparing BPN and RNN approaches.

Open Source

Publications

AgentTelemetry: A Fault Detection Benchmark and Toolkit for LLM Agent Observability

AIware 2026 · 380–387

Benchmark and open-source toolkit for detecting faults in LLM agents: a taxonomy of 14 fault types mapped to 9 agent-specific span kinds, a 490-cell evaluation framework, and a pip-installable library with adapters for seven agent frameworks. Benchmark & Dataset Track, co-located with FSE 2026 in Montreal.

PowerApps Canvas App Accessibility Guidelines

Microsoft whitepaper

Authored guidelines for building accessible applications in Microsoft PowerApps — now a go-to resource influencing how apps are designed and built with accessibility in mind.

Full list on Google Scholar · ORCID · OpenReview · DBLP

Speaking & Reviewing

Deterministic Outcomes in Non-Deterministic Systems: Engineering Reliability for AI Agents

Speaker · Feb 2026

Technical session at an AI Agents event on building production-grade agentic infrastructure by applying distributed-systems principles to LLMs, targeting five-nines reliability for enterprise AI operations.

TelecomReseller Podcast

Media

Featured in "Inside Meta's Work on AI Agent Reliability and Platform Operations."

Conference Reviewing

Reviewer · 10+ papers

Reviewer for IEEE ICCCNT 2025, ICDSNE 2025, and IEEE AIC 2025; Technical Program Committee member for FTNCT 2025. Papers spanning data science, networking, computing, and applied intelligence.

Education

Georgia Institute of Technology

April 2020 – May 2023

M.S. in Computer Science

Specialization in Computing Systems

BITS Pilani

August 2013 – June 2018

M.S. in Mathematics

B.E. in Mechanical Engineering