I build intelligent, scalable, cloud-native systems
combining backend engineering, machine learning, and DevOps.
π Website Β· πΌ LinkedIn Β· βοΈ Email
Application-Level AI Governance Β· SDK + VS Code Extension
π GitHub: https://github.com/samuelbrhane/TokVigil
π Website: https://tokvigil.com
A developer-first AI usage control platform that enforces limits, budgets, and governance policies directly inside application code rather than at the infrastructure or proxy layer.
Key capabilities
- Python and TypeScript SDKs wrapping LLM calls
- Per-user, per-feature, and per-plan usage enforcement
- Per-request cost caps and token limits
- Redis-backed real-time counters
- Deterministic reason codes for blocked requests
- Centralized audit logs and API key tracking
- Environment isolation (dev / staging / prod)
- BYOK support with consistent tracking
- VS Code extension for in-editor usage visibility and policy inspection
Tech stack
Market Forecasting Platform Β· Multi-Horizon Probabilistic Modeling
π GitHub: https://github.com/samuelbrhane/Nordict
π Website: https://nordict.net
A production-grade market forecasting platform delivering multi-horizon probabilistic predictions (24H, 30D, 12W, 12M) with confidence scoring, alerts, and historical forecast-vs-actual analysis.
Key capabilities
- Rolling inference pipelines with hourly and daily refresh cycles
- Multi-horizon probabilistic forecasting with uncertainty estimation
- XGBoost-based modeling with structured evaluation workflows
- Authenticated dashboards for forecasts, alerts, and performance tracking
- Forecast vs actual historical analysis
- Subscription billing with plan-based feature enforcement
- Infrastructure deployed via Terraform with CI/CD automation
Tech stack
Agentic Operational Intelligence Platform Β· Multi-Agent Orchestration
An industrial-grade multi-agent system that bridges technical documentation, live SQL databases, and operational signals to generate actionable mitigation strategies in energy, logistics, and infrastructure environments.
Core architecture
- Agent A (Librarian β RAG): searches manuals and safety SOPs via Kendra/OpenSearch
- Agent B (Analyst β SQL): converts natural language into SQL queries for Athena/RDS
- Agent C (Strategist β Reasoning): synthesizes outputs into structured recommendations
- Orchestrator: AWS Step Functions coordinating agent workflows and escalation logic
Industry-grade capabilities
- Human-in-the-loop approval before high-cost actions
- Cost guardrails integrated with TokVigil
- Multi-agent tool calling via AWS Bedrock
- Secure IAM role-based access with least privilege
- Secrets stored in AWS Secrets Manager
- Infrastructure fully deployed using AWS CDK
Tech stack
- Python
- TypeScript
- JavaScript
- Go
- PostgreSQL
- MySQL
- MariaDB
- Microsoft SQL Server
- MongoDB
- Amazon DynamoDB
- Amazon Aurora
- Amazon S3
- Redis
- pgvector
- FastAPI
- Django
- Node.js
- Express.js
- NestJS
- REST APIs
- WebSockets
- Background workers and async processing
- Go (backend services)
- React
- Next.js
- Component-based UI architecture
- Server-side rendering and performance optimization
- TensorFlow
- PyTorch
- Scikit-learn
- XGBoost
- MLflow
- Pandas
- NumPy
- Time-series forecasting and regression models
- AWS (EC2, S3, RDS, ECS, Lambda, SageMaker)
- Docker
- Kubernetes
- GitHub Actions
- Infrastructure as Code (Terraform)
- Monitoring, logging, and cost optimization
- End-to-end systems from idea to architecture to production
- Scalable, cloud-ready, AI-powered platforms
- Clean, maintainable, and well-documented code
- Reliable execution and clear communication