For teams hiring a Senior AI Engineer
I ship LLM systems with the limits written down.
Retrieval, prompt safety, cost controls, fallbacks and evaluation, built into a production backend by one engineer. Then measured, and the gaps stated.
What you get
- 5.1 → 1.2points of grading spread, after a three-round eval I designed and ran
- 670+tests on an AI backend I built alone
- 80,000token budget per learner per module, enforced server-side
- $0model cost per view for 4,201 generated career guides
The three artifacts to read
-
Hashflag Skills
An eval with a target set in advance, fixtures built to break the grader, and a record of where the model was wrong.
-
J0DI3 AI backend
FastAPI and pgvector RAG with circuit breakers, token budgets, prompt-injection defenses and output guardrails.
-
MOS Translator
A schema-validated generation pipeline with tiered data fallback and no runtime model dependency.
How I work with models
- Deterministic code computes anything a user will act on. The model does the reading and has to quote its evidence.
- The failure mode is chosen before the feature is built: fail open, or stop.
- Prompts live in the repository and are reviewed like code.
- What has not been evaluated is stated next to what has.
The detail is on How I Build and in the decision records.
Enterprise AI experience
As a Software Engineer II at Microsoft from March 2021 to January 2025, I built AI systems for enterprise customers: a real-time speech-to-text system for Starbucks on Azure OpenAI and Azure Speech Services with PII redaction, and a full-stack RAG chat application for Domino’s Pizza, evaluated with RAGAS. That work belongs to Microsoft and its customers, so there is no repository to link. I am glad to talk through it.
Skills
Logistics
I am based in Atlanta, Georgia, and work remotely. I am open to Senior AI Engineer roles.