Lagos, Nigeria · operational since 2026

Operational intelligence
for African infrastructure.

PHANG builds systems that eliminate operational blindness, trust failures, and fragmented execution in markets where those failures have the highest cost. Each product is an infrastructure layer — not an interface. The moat is what the system knows, enforces, and remembers.

Classification
Operational Intelligence
Systems live
2 production
Infrastructure
Single-tenant · self-hosted
AI layer
DeepSeek · self-orchestrated
Live
InfoPro
infopro.ng
failure mode: manual interpretation at scale degrades faster than data volume grows

Document intelligence and analysis pipeline for institutional workflows. Replaces fragile manual interpretation with structured reasoning over unstructured data — CSV analysis, document intelligence, AI-assisted writing. Designed for operators who process high volumes of documents under time and accuracy constraints.

document-intelligence analysis-pipeline Go · Next.js DeepSeek
Live
NextDoor NG
nextdoorhome.site
failure mode: information asymmetry in Lagos housing markets transfers cost and risk entirely to tenants

Trust and coordination infrastructure for fragmented African housing markets. Enforces full cost disclosure, agent identity verification, and complaint escalation at the system level — not as policy, but as architecture. WhatsApp-first search for markets where web friction is a real barrier.

trust-infrastructure agent-verification FastAPI · Python WhatsApp-native
Internal
MyTodo
strategic execution layer
failure mode: human execution systems collapse when assumptions break and no consequence modeling exists

Adaptive strategic intelligence for long-term human execution systems. Models consequences, detects failure cascades before they compound, and rebuilds execution plans when environmental conditions change. Not a task manager. An execution environment with memory and consequence awareness.

adaptive-intelligence consequence-modeling stateful-memory strategic-execution
Internal
PHANG CUA
operational command layer
failure mode: AI-native organizations have no runtime observability — incidents are discovered from users, not systems

Operational command interface for the PHANG infrastructure layer. Two-layer architecture: deterministic slash commands for known operations, DeepSeek reasoning for diagnostic queries. Docker socket access, event telemetry, container lifecycle control — all through Telegram.

operational-observability command-interface Python · asyncio Telegram
01

Single-tenant, self-hosted

One VPS. One PostgreSQL instance shared across products. One gateway routing all traffic. No managed services that abstract operational visibility away from the operator. The system is owned end-to-end.

02

Unified event telemetry

All products emit to a shared ops_events table. System state is observable from a single query surface. The CUA reads this table to reason about infrastructure health without external monitoring dependencies.

03

AI layer — DeepSeek, self-orchestrated

No OpenAI. No Anthropic dependency for production inference. DeepSeek is called directly from product code with deterministic context injection. The AI layer reasons over real operational data, not hypothetical inputs.

04

REUSE → EXTEND → CONSOLIDATE

New infrastructure is only introduced after a seven-question reuse audit confirms no existing component can be extended. Complexity is not added to signal maturity. Four percent RAM utilization under real load is the target, not the floor.

05

WhatsApp as a first-class interface

In markets where smartphone penetration outpaces web adoption, WhatsApp is not a notification channel — it is the primary interaction surface. PHANG systems are designed for this reality, not retrofitted to it.

06

Trust at the architecture level

Trust mechanisms are structural, not cosmetic. Agent verification gates listing creation. Cost disclosure is enforced by the data model, not by policy pages. Complaint routing is wired into the schema. The system cannot be configured to bypass these constraints.

Gateway
api.infopro.ng
Stack
Go · Python · Next.js
AI inference
DeepSeek
Hosting
Hetzner CX22 · self-managed
Database
PostgreSQL 16 · shared
Base
Lagos, Nigeria
Trust as architecture
Trust is not a feature you add after the product ships. It is either in the schema, the verification flow, and the enforcement logic — or it does not exist. Every PHANG system is designed so that the unsafe path is architecturally harder than the safe one.
Operational visibility first
A system you cannot observe is a system you cannot govern. Before a new capability is added, the question is: will we know when it fails? Unified event telemetry, structured logging, and runtime command access are not features — they are the baseline.
Adaptive, not static
Static systems break at the edges where real decisions happen. PHANG systems recalculate under changed conditions — new data updates recommendations, broken assumptions trigger plan reconstruction, new inputs modify routing. A system that cannot adapt is a snapshot, not infrastructure.
Consolidation before complexity
One VPS. One database. One gateway. Every architectural decision is evaluated against REUSE → EXTEND → CONSOLIDATE before new infrastructure is introduced. Operational maturity is running real production workloads efficiently — not the number of services in the compose file.
Market-first design
African markets are not underserved versions of Western markets. They have distinct infrastructure constraints, trust patterns, and interaction surfaces. Systems designed for these realities from the start are structurally different from systems retrofitted to them. That difference compounds.