"Outcomes over demos. Governance over hype."
I modernize brittle, legacy/no‑code automations into reliable, scalable, cost‑efficient systems. I don't just write scripts; I build production-ready architectures with binary acceptance tests, P95 latency targets, and live ROI dashboards.
I bridge the gap between AI research and Enterprise reliability.
| Category | Technologies |
|---|---|
| Language |   |
| AI Orchestration |   RAG (Graph/Vector) |
| Backend |   Pydantic |
| Data & Vector |    |
| Infrastructure |   GitHub Actions |
I build systems that hit specific Service Level Objectives (SLOs).
Automated reconciliation with 100% duplicate detection.
- Scope: Docling OCR + Pydantic validators + Anomaly digests.
- Metric: ≥98% field accuracy on 200‑doc test sets.
- Stack: Python, FastAPI, GraphRAG.
Email/WhatsApp resolution with strict governance.
- Scope: Citations required, QA gating, Sentiment routing.
- Metric: P95 response < 2 minutes; Breach alerts < 30s.
- Stack: LangGraph, Redis Queues, LiteLLM.
Idempotent calendar operations.
- Scope: Intake → Qualifier → Slot Picker → Reminders.
- Metric: +20% show‑rate vs baseline; < 60s write latency.
- Predictability: Typed data flows end‑to‑end (TypeScript/Zod/Pydantic). No "stringly typed" code.
- Security: Least privilege, audit logs, and PII redaction by default.
- Observability: If it isn't logged in Langfuse/Phoenix, it didn't happen.
- Ownership: Code‑first, no lock‑in. I build systems your team can extend.
I work best with ops‑minded founders who value clear scope and sustainable systems over throwaway prototypes.
- 10‑Day Modernization Audit: Latency/cost baselines, SLAs, and a fixed pilot SOW.
- Pilot Build (10–14 days): Pass/Fail delivery based on acceptance criteria.
- Ongoing Ops: Monthly SLOs and change-managed improvements.
