AI / LLM Engineer building production AI for financial services — credit-risk models, multi-agent pipelines, and the secure ML platforms that ship them. 5+ years from data engineering to ML and cloud architecture, now focused on applied LLM systems. AWS ML Specialty certified.
| Certification | Issuer |
|---|---|
| AWS Certified Machine Learning – Specialty | Amazon Web Services |
| AWS Certified Cloud Practitioner | Amazon Web Services |
| GCP Professional Data Engineer | Google Cloud |
| Snowflake SnowPro Advanced: Data Engineer | Snowflake |
| dbt Analytics Engineering Certification | dbt Labs |
| Domain | Technologies |
|---|---|
| LLM Engineering | LangChain · LangGraph · RAG · MCP · Prompt Engineering · OWASP LLM Top 10 · Semantic Caching · LLM Gateways |
| AgentOps | Multi-Agent Systems · LangGraph · CrewAI · Tool Use · Agent Memory · Autonomous Coding Agents · n8n |
| MLOps / LLMOps | MLflow · Feast · Airflow · Evidently AI · RAGAS · Prompt Versioning · A/B Testing · Model Registry |
| DevSecOps | GitHub Actions · ArgoCD · Crossplane · OPA Gatekeeper · Trivy · Semgrep · Cosign · SBOM · SLSA |
| Data Engineering | Apache Kafka · Apache Spark · dbt · Redis · Airflow · SQL lineage · ETL/ELT · PostgreSQL |
| ML / Deep Learning | PyTorch · XGBoost · scikit-learn · HuggingFace · Federated Learning · Differential Privacy · Fairlearn |
| Cloud & Infra | AWS (EKS · SageMaker · Redshift · S3 · IAM · CloudWatch) · GCP · Terraform · Docker · Kubernetes |
| Security & Governance | ECOA / Fair Lending · GDPR · PCI-DSS · SOC2 · Model Cards · Audit Trails · Data Lineage |
| Repo | What it does | Stack |
|---|---|---|
agent-autopsy |
Framework-agnostic agent interceptor — builds execution flame graphs, detects failure patterns (loops, token storms, goal drift), generates structured post-mortems | Python · networkx · FastAPI |
llm-gateway |
Production LLM gateway: OpenAI-compatible API, semantic caching, multi-model routing, cost analytics dashboard | FastAPI · Redis · sentence-transformers |
llm-safety-auditor |
250+ adversarial attack payloads, OWASP LLM Top 10 scoring, red-team report generation, Streamlit dashboard | FastAPI · sentence-transformers · reportlab |
nano-finbert |
Tiny transformer (~2M params) trained from scratch on financial text — outputs structured market signals: sentiment, entities, sectors, event type, impact score | PyTorch · FastAPI |
| Repo | What it does | Stack |
|---|---|---|
autonomous-coding-agent |
Reads GitHub issues → understands codebase → generates fix → runs tests → opens PR | Claude · LangGraph · PyGitHub · FastAPI |
mcp-diagram-agent |
MCP server: describe any system → production-ready Excalidraw architecture diagram via Claude. Fully typed tool surface, 97%+ coverage | MCP · Claude · FastAPI · mypy |
finance-agent-crew |
Multi-agent financial intelligence: SEC EDGAR + earnings calls + news → investment brief | LangGraph · Claude · n8n · FastAPI |
ops-autopilot |
Autonomous SRE agent: receives alerts → LLM root-cause diagnosis → creates GitHub fix PR + Slack summary + runbook entry | FastAPI · Anthropic · PyGithub |
multi-agent-research-crew |
LangGraph crew (researcher + analyst + writer + critic) producing structured research reports with SSE streaming progress | LangGraph · Claude · FastAPI |
| Repo | What it does | Stack |
|---|---|---|
llmops-eval-platform |
RAGAS RAG evaluation + Claude-as-judge + A/B testing + prompt versioning + cost tracking | Python · RAGAS · FastAPI · SQLAlchemy |
nvidia-nim-rag-techniques |
5 production RAG techniques on NVIDIA NIM: hybrid search/RRF, cross-encoder reranking, query rewriting (HyDE/multi-query/step-back), context compression, corrective RAG | NVIDIA NIM · LangGraph · FastAPI |
mlops-feature-platform |
Enterprise MLOps for financial ML: Feast feature store + MLflow registry + Airflow pipelines + Evidently AI monitoring | Feast · MLflow · Airflow · Evidently |
kafka-stream-feature-store |
Real-time ML feature store: Kafka → Redis pipeline with sub-60s feature freshness for low-latency inference | Kafka · Redis · FastAPI |
financial-model-governance |
ML governance for finance: ECOA fairness testing, disparate-impact analysis, model cards, audit trails, approval workflows | Python · ECOA · Model Cards |
| Repo | What it does | Stack |
|---|---|---|
fintech-devsecops-pipeline |
End-to-end DevSecOps: SBOM generation, Trivy scanning, OPA policy enforcement, Cosign image signing, SLSA Level 3 provenance | GitHub Actions · Trivy · Cosign · OPA · ArgoCD |
k8s-platform-engineering |
Internal Developer Platform: ArgoCD GitOps + Crossplane infra provisioning + OPA Gatekeeper policies + golden-path templates | Kubernetes · ArgoCD · Crossplane · OPA |
adaptive-cognitive-rag |
RAG that measures cognitive state from behavioral signals and adapts explanation depth; builds a per-session knowledge graph of mastery gaps | RAG · Knowledge Graph · LangChain · FastAPI |
Full portfolio spans computer vision, AIOps, n8n automation, Kubernetes platform engineering, RAG assistants — all with CI/CD, Docker, 95%+ test coverage.
Private repos available on request — federated learning for credit risk, LLM red-teaming, on-device LLM optimization, SQL-to-DAG compilers. Contact via LinkedIn.
Production-grade repos · CI/CD green · 95%+ test coverage · Docker · AWS · nagizaaz.vercel.app · LinkedIn