Recruiter Brief¶
Recruiter screening view · 90 seconds
MLOps engineer who builds systems that survive production¶
Two years of focused ML engineering on top of 14 years of operations leadership. Three production services, three incidents diagnosed from first principles, one open-source template encoding everything learned. The portfolio makes the evidence reviewable.
Best-Fit Roles¶
Primary fit
MLOps Engineer¶
Serving, CI/CD, monitoring, MLflow hygiene and deployment reliability for ML systems in production — the discipline the entire portfolio is built around.
Equally strong
ML Engineer¶
Applied ML roles where model work needs APIs, testing, deployment and a clear handoff into an engineering workflow — not just a notebook.
Strong second
Data Scientist (production-leaning)¶
DS roles where the model is not the end state — serving, monitoring and operability matter alongside the modeling work.
Adjacent path
Data / AI Platform Engineer¶
Pipeline, batch and agentic-systems roles connected to production ML — PySpark, validation, feature engineering, governed AI workflows.
What Makes Me Different¶
2026 differentiator
Governed AI-assisted development¶
I engineer my AI workflow instead of hiding it: an AUTO/CONSULT/STOP behavior protocol, an audit trail and eval gates — codified in immutable Python, not a README promise.
Debugging
I measure before I guess¶
Three production incidents, three root causes found by measuring, not trial and error. The clearest: an 81% error rate under load traced to CPU contention and fixed to 0% — with CPU cost cut in half.
Ownership
Decisions survive the handoff¶
18 ADRs across the portfolio, 38 more in the Production Template — runbooks and status pages so another engineer can operate the system, not just read about it.
Career arc
Operations background is a feature¶
14 years of budgets, vendors, deadlines and team pressure. Cost discipline, scope judgment and documentation habits come from real stakes — not a certification module.
First 90 Days Contribution¶
Days 1-30
Ship fast, break nothing¶
Run the stack locally, learn the model lifecycle and deployment path, and close the first small gaps — onboarding friction, flaky tests, missing docs.
Days 31-60
Take real delivery load¶
Own FastAPI endpoints, validation checks, MLflow hygiene, CI/CD tasks, Docker/Kubernetes artifacts or monitoring improvements — reviewed, then trusted.
Days 61-90
Own a reliability win end-to-end¶
Take one scoped improvement from issue to shipped documentation: smoke tests, readiness checks, drift alerts, runbooks or deployment evidence — the same rigor as the 81%→0% fix, applied to your stack.
What To Look For In The Portfolio¶
Project judgment
A template, not just projects¶
The strongest signal is the Production Template: 38 encoded anti-patterns and 43 ADRs turning three projects' worth of pain into a reusable, governed starting point other engineers can build on.
Debugging ability
Incidents, not just metrics¶
Load testing, inference-path debugging, a caught data-leakage bug before it hit a published metric — the portfolio shows the failures as clearly as the wins.
Communication
Written for your whole team¶
Architecture notes, model cards, runbooks and deployment evidence written so both engineers and non-technical stakeholders follow the story without a walkthrough.
What I Am Building Next¶
Live evidence
More real traffic windows¶
Short, cost-controlled live demos to capture fresh Grafana, Prometheus and MLflow evidence without leaving infrastructure online permanently.
Frontier
Agentic systems, same governance¶
agent-local extends the template's AUTO/CONSULT/STOP philosophy
to local, multi-tier LLM agents — grammar-constrained routing already
passing its quality gate 20/20.
Collaboration
More public review signals¶
External feedback, PR review examples or open-source contributions so the portfolio shows how I work with other engineers, not only alone.
Domain fit
A project closer to operations¶
Inventory, staffing, cost anomalies or operations forecasting — where 14 years of business context is a direct modeling advantage, not just a bio line.
Context For The Screening Call¶
Runtime evidence
Load tests today, sustained traffic next¶
Current evidence comes from controlled load tests and live development windows. The portfolio captures the behavior; sustained production traffic on a team's system is the natural next chapter.
Cloud infrastructure
GCP and AWS, Kubernetes-first¶
Deep, hands-on GKE, EKS, Terraform and kubectl across three deployed services. Managed ML platforms (SageMaker, Vertex AI, Azure ML) — architecture is familiar; hands-on integration is the fastest onboarding item, not a blind spot.
ML positioning
Engineering-first, not research-first¶
I am not positioning as an ML researcher. My edge is turning applied models into testable, observable, operable systems — the part most teams are actually short on.
Collaboration signals
Solo work with team habits¶
All projects are solo-built with team-operation standards: ADRs, runbooks, model cards, incident writeups. An open-source contribution or external PR review is the next proof point in progress.