Blog

Long-form essays on systems, architecture, and ML in production.

Sep 23, 2026 Why your flawless code will still break during a live demo
Sep 8, 2026 Longer context windows don't eliminate the need for retrieval
Aug 25, 2026 Every layer of configurability has an ongoing maintenance cost
Aug 18, 2026 MLOps in a Regulated Industry: What Actually Changes
Aug 4, 2026 Graceful shutdown handling prevents more incidents than any autoscaling tweak
Jul 21, 2026 Consent state is data too, and it needs the same residency rules
Jul 7, 2026 Cross-region calls in the hot path are a design decision, not an accident
Jun 23, 2026 Shadow deploy new models before they get real traffic
Jun 10, 2026 Evaluating GenAI Guardrail Tooling: What We Actually Use
Jun 9, 2026 Secrets rotation should be routine, not an incident response step
May 26, 2026 An alert that doesn't require action shouldn't page anyone
May 12, 2026 Know your database's real connection limit before you scale horizontally
Apr 28, 2026 Evaluation sets need maintenance too
Apr 16, 2026 What SOC2 Audits Actually Check For, As an Engineer
Apr 14, 2026 A schema registry pays for itself the first cross-team incident it prevents
Mar 31, 2026 Give every service in a chain its own latency budget
Mar 17, 2026 Treat any user-supplied text reaching an LLM as untrusted input
Mar 5, 2026 Debugging Distributed Systems: A War Story
Mar 3, 2026 Blue-green deploys need real thought for stateful services
Feb 17, 2026 Documentation has a half-life, budget for refreshing it
Feb 3, 2026 Rate limiting per tenant, not just globally, prevents noisy neighbors
Jan 27, 2026 Terraform Patterns for Multi-Tenant Infrastructure
Jan 20, 2026 Every model deploy needs a rollback plan as concrete as a code deploy's
Jan 6, 2026 Cloud cost attribution should be automatic, not a spreadsheet
Dec 23, 2025 Severity levels only work if everyone agrees on the definitions
Dec 15, 2025 gRPC vs. REST for Internal Services: What Actually Decided It
Dec 9, 2025 Treat the context window like a scarce resource
Nov 25, 2025 Load test with realistic traffic shapes, not just volume
Nov 12, 2025 API Gateway Patterns for Multi-Region Compliance
Nov 11, 2025 A service with no clear owner accumulates the worst tech debt
Oct 28, 2025 Synthetic data is a supplement, not a substitute
Oct 21, 2025 Understanding PCA Before You Reach for Deep Learning
Oct 14, 2025 Audit unused indexes as carefully as missing ones
Sep 30, 2025 Embeddings drift when the underlying model changes silently
Sep 23, 2025 Evaluating LLM Output Without a Golden Dataset
Sep 23, 2025 Feature Stores: Do You Actually Need One?
Sep 16, 2025 Quarterly access reviews catch what onboarding misses
Sep 9, 2025 Where an LLM Layer Actually Helped a Recommendation Engine
Sep 2, 2025 A dead letter queue nobody monitors is just a queue
Aug 19, 2025 Track token usage like you'd track any other cost center
Aug 19, 2025 Building a Recommendation Engine from Scratch: What We Underestimated
Aug 5, 2025 Backpressure is a feature, not a failure mode
Jul 30, 2025 Lessons from Moving 10+ TB Through a Redshift Pipeline
Jul 22, 2025 A runbook nobody has tested is a guess with formatting
Jul 8, 2025 Vector index choice matters less than metadata filtering
Jun 25, 2025 Fine-Tuning vs. RAG: A Practical Decision Framework
Jun 24, 2025 Canary deploys catch what staging can't
Jun 10, 2025 Assume PII will end up in logs unless you actively prevent it
May 27, 2025 Design for the dependency being down, not just slow
May 14, 2025 Guardrails for LLM Applications in a Pharma Context
May 13, 2025 Detecting hallucinations is a retrieval problem, not just a model problem
Apr 29, 2025 Connection pool exhaustion looks like everything else
Apr 15, 2025 Prompts are code and need the same review process
Apr 1, 2025 Retries without jitter cause the outage they're meant to prevent
Mar 21, 2025 Prompt Engineering for Internal Tools Is a Different Discipline
Mar 18, 2025 Write the model card before you need to defend the model
Mar 11, 2025 Building a Self-Serve Analytics Layer, and What Breaks First
Mar 4, 2025 Data contracts between teams need a version, like an API
Feb 19, 2025 Timeouts don't compose the way you would hope
Feb 5, 2025 Chunking strategy matters more than the embedding model
Jan 22, 2025 If everything is logged at ERROR, nothing is
Jan 8, 2025 Feature flags need an expiry date
Dec 20, 2024 A postmortem that names a person has already failed
Dec 9, 2024 Migrations should be boring
Nov 25, 2024 SageMaker cold starts add up fast
Nov 19, 2024 Hexagonal Architecture, Six Months In
Nov 11, 2024 Idempotency keys are cheap insurance
Oct 29, 2024 Scope tenancy at the data layer
Oct 15, 2024 Add observability before you need it
Oct 2, 2024 Version the contract, not the code
Sep 23, 2024 What ClickHouse Taught Me About Read-Heavy Systems
Sep 23, 2024 Multi-Tenant Data Isolation: The Boring Problem That Isn't
Sep 20, 2024 Always plan before apply, no exceptions