Md Yunus
DevOps/SRE | 8 PRs merged to Cloudflare & Docker | OCI Certified | Building infrastructure that ships
New Delhi
1+ year of work experience
Jul 20262 mosPresent
Remote
Designing reproducible Terminal-Bench tasks that evaluate LLMs on real-world Linux, DevOps, debugging, and automation workflows using Docker, Python, Bash, and deterministic tests.
May 20264 mosPresent
Remote
Oracle ACE Associate — promoted from ACE Apprentice (Jan–May 2026) for demonstrated technical depth in OCI.
- Designed 12+ OCI reference architectures spanning Compute, VCN networking, IAM policy frameworks, and DevOps pipeline services — published for Oracle's developer community and used in customer implementations
- Evaluated cost-performance trade-offs across managed services (OKE, Autonomous DB) vs. self-managed infrastructure, providing decision frameworks for 5+ deployment patterns
- Published technical content reaching 2,400+ followers on cloud-native OCI workload patterns
Certifications: OCI 2025 Architect Associate · OCI DevOps Professional · OCI Developer Professional
Oct 202511 mosPresent
New Delhi, Delhi, India
Cloudflare — Vinext Framework (github.com/cloudflare/vinext)
→ 6 merged PRs (#258, #345, #249, #197, #193, #151): Refactored shared server request handling, fixed ESM module resolution breaking production builds, improved CLI config loading, added snapshot tests for entry template generators, resolved Pages Router context and export issues
→ Contributions running in Cloudflare's edge infrastructure serving millions of developers globally
Docker — docker-agent (github.com/docker/docker-agent)
→ PR #1973: Fixed context window overflow with auto-recovery and proactive compaction—added 33 comprehensive tests covering edge cases and structured error messages, preventing infinite conversation crashes
→ PR #2175: Fixed TUI table text copy misalignment by mapping visual columns to rune indices; simplified parameters and resolved lint warnings
n8n — Workflow Automation Platform
→ Contributed Docker Compose improvements and debugging enhancements for self-hosted deployments, improving developer experience for 50,000+ self-hosted users
Hacktoberfest Recognition
→ Completed 2023, 2024, and 2025 cycles—recognized in official Hall of Fame
→ 70% PR acceptance rate (17 of 24 merged) across diverse open-source ecosystems
→ 10 PRs submitted in 2025 cycle; 7 accepted by maintainers with focus on code quality, bug fixes, and documentation
Contribution Distribution: 53% commits, 43% issues/discussions, 4% pull requests, 1% code reviews—demonstrating deep project engagement beyond code submissions
Jun 20262 mosJul 2026
Dadri, Uttar Pradesh, India
Four weeks inside a 1,840 MW power station learning what reliability engineering looks like when the blast radius is measured in megawatts, not error rates.
Worked under Shri A. K. Dhiman, DGM (C&I), studying the Control & Instrumentation systems that keep NTPC Dadri online.
What I mapped to SRE/DevOps:
→ Active/standby with dual network paths — DDCMIS controllers run in redundant pairs on independent A/B LANs. Continuity is designed in, not bolted on.
→ Observability by convention — 4 mA = valid zero; 0 mA = broken circuit. A failure cannot masquerade as a healthy reading. No dashboards needed to tell the difference.
→ Game days, hardware edition — The ATT SOP tests one turbine protection device at a time, proves the backup trip path healthy first, then restores before moving to the next. Verify the fallback before you test the failure.
→ Change control without a pipeline — DDCMIS LAN access, engineering changes, and patching follow gated, documented processes on a Purdue-zoned, firewalled network. Controlled change enforced by procedure, not tooling.
Documented the full training in a 39-page report with 36 figures across 10 chapters: thermal process, instrumentation, closed-loop control, DDCMIS architecture, turbine protection, OT security, calibration, and loop checking.
SRE didn't invent redundancy, observability, or controlled change. Industrial control systems have been solving the same reliability problems for decades — just with physical consequences.
Oct 20252 mosNov 2025
Remote
End-to-end ownership of an ML system for aviation operations—from 2.5M+ raw records to a deployed, tested, containerized production service built on San Francisco International Airport (SFO) data.
→ Processed and aggregated 2.5M+ aviation time-series records with pandas/numpy; cleaned noise to improve downstream model reliability and transformed raw data into daily KPIs for operational analysis
→ Built Prophet-based 12-month landing volume forecasts to support capacity and staffing planning; modeled long-term trends via linear regression providing actionable insights for resource allocation
→ Implemented Isolation Forest anomaly detection for early identification of abnormal landing patterns—flagged irregular days 3 weeks faster than manual spreadsheet analysis, enabling proactive operational response
→ Delivered a typed FastAPI backend serving forecasts, trends, and anomaly insights; built interactive Dash + Plotly dashboard for real-time decision support used by capacity planning teams
→ Containerized the full system with Docker; automated testing and deployment via GitHub Actions CI/CD—100% pass rate across 9 unit/integration tests with pytest, reducing deployment time from 2 hours manual process to 12-minute automated pipeline
Stack: Python 3.11 · pandas · numpy · Prophet · scikit-learn · FastAPI · Dash · Plotly · Docker · GitHub Actions · pytest
Md’s AI Stack
Education
Bachelor of Technology - BTech, Electronics and Communications Engineering
Guru Nanak Dev University, Amritsar
20232023
Aspire Leaders Program 2025, Leadership & Social Impact
Aspire Institute
20252025
Join Md and 10k+ other professionals on
Tal is the app where bosses meet engineers & builders directly, without middlemen
Download Tal Boss