Piyush Chauhan pxyushch

🚀 Core Engineering Radar

                      [ FRONTIER MODEL EVALUATIONS ]
                                    ▲
                                    │
               ┌────────────────────┴────────────────────┐
               │                                         │
   [ CONTAINER ISOLATION ]                     [ CLOUD AUTOMATION ]
   • Docker cgroups (Strict 1GB RAM)           • GitHub Actions Matrix Sharding
   • Non-root runtime isolation                • GCP Compute Engine & Cloud Run
   • Deterministic OOM signal trap             • Automated healthcheck rollbacks
               │                                         │
               └────────────────────┬────────────────────┘
                                    │
                                    ▼
                      [ SYSTEMIC FAILURE TAXONOMY ]
                        (Wildcards, RBAC Drift)

⚡ Tech Stack & Capabilities

Primary Languages & Systems Engineering

DevOps, Cloud, Sandboxing & Testing

AI, ML & Interpretability Frameworks

🔬 Benchmark Engineering & Real-World Specialization

At Ethara.ai, I design deterministic evaluation harnesses to rigorously test frontier LLMs on enterprise cloud infrastructure tasks. Rather than relying on simple string-matching, our benchmarks run in isolated Linux containers with multi-axis rubric verification:

🎯 Deterministic Benchmarking & Task Authoring: Engineering reproducible task environments across Docker, Kubernetes, Helm, Terraform, and CI/CD with unit tests and golden solutions.
🐳 Hermetic Sandbox Execution: Isolating model-generated scripts inside ephemeral Docker runners with restricted network egress, non-root execution, and strict memory cgroups.
🚨 Systemic AI Mistake Taxonomy: Cataloging recurring model flaws—such as unpinned base image digests, excessive * permissions in Kubernetes roles, and hardcoded cloud secrets.
☁️ Cloud CI/CD & Infrastructure Automation: Architecting production deployment workflows with GitHub Actions, Google Cloud Platform (Compute Engine, GCR, Cloud Storage), and automated health-check rollback gates.
🧠 Explainable AI Research (DepreXAI): Co-authored clinical NLP research utilizing transformer ensembles (PyTorch, Hugging Face) paired with SHAP/LIME feature attribution (94.2% accuracy, 0.91 macro F1).

📊 Evaluation Matrix & Benchmark Results

Pass rates and deterministic failure profiles observed across frontier models in our sandboxed test suites:

Benchmark Domain	Target Models	Core Verification Standard	Caught Failure Modes	Pass Rate
Multi-Stage Dockerfiles	Claude 3.5 Sonnet, Codex, GLM-4	Non-root user, minimal base image, multi-stage build	Default root user, cache invalidation, bloated layers	`78.4%`
Kubernetes RBAC & NetPol	Claude 3.5 Sonnet, GLM-4, DeepSeek	Namespace isolation, explicit verbs, no cluster-admin	Wildcard `*` verbs, missing egress boundaries	`64.2%`
Terraform GCP Infrastructure	Claude 3.5 Sonnet, Codex	Resource naming conventions, IAM least privilege	Hardcoded secrets, unencrypted bucket configs	`71.8%`
Production Helm Charts	Codex, GLM-4, DeepSeek	Values schema validation, probe health configurations	Missing readiness probes, template linting errors	`62.0%`
CI/CD Pipeline Security	Claude 3.5 Sonnet, Codex	GitHub Actions token permissions, step timeouts	Unpinned third-party actions, excessive GITHUB_TOKEN	`81.5%`

📋 Flagship Repositories & Engineering Suites

Repository	Focus & Domain	Architecture & Stack	Verification Method
`devops-llm-benchmark`	DevOps LLM Benchmark Suite	Docker • Kubernetes • Terraform • CI/CD	Docker Sandboxing + Pytest
`cloud-cicd-gcp-pipeline`	Cloud CI/CD Automation	GitHub Actions • Docker • GCP	GCP Deployment + Live Healthchecks
`playwright-qa-automation`	Enterprise E2E Test Suite	Playwright • Python • Allure Reports	Parallel Matrix Shards (4x Speed)
`deprexai`	Explainable Clinical NLP	PyTorch • Transformers • SHAP • LIME	Cross-Validation (94.2% Acc, 0.91 F1)
`Automation-Testing-tool`	QA Test Automation Framework	Python • Playwright • AstraQ	Deterministic Selector Assertions
`taskflow-rbac-app`	Full-Stack Enterprise RBAC	React • Express • Prisma • SQLite	Automated Route Bypass & Audit Tests

🏆 GitHub Official Milestone Achievements

Verified milestones awarded for open-source contributions, technical discussions, and code reviews.

Badge	Distinction	Verification Metric	Status
🧠 Galaxy Brain	Bronze (x1)	2+ accepted answers to technical Q&A discussions	`UNLOCKED`
🪂 YOLO	Unlocked	Merged pull request without blocking review requirements	`UNLOCKED`
🦈 Pull Shark	Silver (x2)	16+ merged pull requests into default branches	`UNLOCKED`
⚡ Quickdraw	Unlocked	Issue/PR closed within 5 minutes of opening	`UNLOCKED`
👥 Pair Extraordinaire	Bronze (x1)	Merged co-authored commits via pull requests	`UNLOCKED`

📈 Activity Telemetry & Engineering Rank

📬 Open for Engagements

role: "DevOps Engineer & AI Evaluation Specialist"
current_company: "Ethara.ai"
domains:
  - "Frontier Model Benchmarking (Claude 3.5 Sonnet, Codex, GLM-4, DeepSeek)"
  - "Deterministic Cloud Sandboxing & Ephemeral Test Containers"
  - "Automated CI/CD Pipelines & Kubernetes Security Verification"
contact:
  email: "peeyushrajput321@gmail.com"
  linkedin: "https://linkedin.com/in/piyush-chauhan-8a02a0251"
  github: "https://github.com/pxyushch"
status: "Available for high-impact AI Evaluation & DevOps opportunities"

Navigation Menu