Open to consulting engagements and select full-time roles

Christian Chukwuka

AI Systems Engineer & Founder, ChuksForge AI Solutions — Lagos, Nigeria

I design and ship production AI systems, including LLM pipelines, multi-agent architectures, and the SaaS platforms built around them, for businesses that need them to hold up under real use, not just demo well. BOS, my flagship product, runs live for Nigerian small business owners today.

A background in systematic trading shapes how I build: measure everything, favor signal over noise, and design for the conditions you didn't plan for.

See the work → GitHub
Christian Chukwuka
01 — In production
BOS
WhatsApp business OS, live with Nigerian MSMEs
02 — Open-core
EvalCI
LLM regression testing, design-partner stage
03 — Deployed
ChuksForge FinOps
Ledger & billing infrastructure for AI companies
About

Engineer by trade.
Trader by discipline.

I'm Christian Chukwuka, founder of ChuksForge AI Solutions Ltd, a registered AI engineering consultancy in Lagos, Nigeria. I design and deploy production-grade AI systems, including end-to-end LLM applications, evaluation pipelines, and multi-tenant SaaS backends, with an emphasis on reliability, scalability, and real-world usability.

My background in systematic trading across forex and crypto shapes how I approach engineering: prioritizing signal over noise, building for robustness under uncertainty, and relying on measurement over intuition.

Several projects below are confidential commercial engagements; the code and architecture stay private, but the standard is the same: production quality, not proof-of-concept. I also maintain ForgeObserver, an internal LLM observability platform used across my own systems.

cafe shot
AI & LLM Systems
  • LLM application developmentProduction pipelines, not demos
  • Agentic systemsMulti-agent orchestration
  • RAG & knowledge systemsRetrieval-grounded reasoning
  • Evaluation & benchmarkingRegression-safe LLM behavior
Platform & Product Engineering
  • Multi-tenant SaaS architectureFrom schema to billing
  • Production hardeningScale audits, incident response
  • API security & proxyingAuth, rate limiting, isolation
  • Monorepo architectureTurborepo, shared packages
Tools

Working stack

Languages & Frameworks
Python · FastAPI · Next.js · React · Node.js · Turborepo · Tailwind CSS
AI & LLM
LangGraph · LangChain · OpenAI · Anthropic Claude · Gemini · OpenRouter · Mistral · RAG pipelines
Infra & Data
PostgreSQL · Supabase · Redis · Celery · Kubernetes · Railway · Prisma · WhatsApp Business API
Featured work

Products & projects

Click a project to expand architecture, contributions, and current status.

01
Live Confidential engagement FastAPI
BOS — WhatsApp Business Operating System
A business OS for Nigerian MSMEs
+
BOS dashboard/chat screenshot
Overview

A merchant messages BOS the way they already message customers, and it records sales, tracks expenses and debtors, manages inventory, and sends daily business briefings automatically. It has grown into a full financial back office: PDF invoicing and device-sale receipts generated from a single message, staff and multi-branch support, bulk debtor clearing, and analytics. Plus, more recently, a companion mobile and web app built on the same backend.

Key contributions
  • Shipped the product and owned production hardening pre-launch. Fixed a cross-region infrastructure misconfiguration that cut latency roughly 4x, plus rate limiting, dead-letter queues, and compliance work
  • Rebuilt the natural-language parser architecture and hardened it against real Nigerian merchant phrasing across many rounds of review
  • Shipped invoicing, device-sale receipts, staff/branch management, bulk debt clearing, and analytics
  • Extended the product into a full React Native / Expo mobile and web app, reusing the existing FastAPI backend end-to-end
  • Diagnosed two separate production incidents to root cause, each closed out with a regression test
Stack
FastAPICeleryPostgreSQL RedisWhatsApp Business API React Native / ExpoRailway

Confidential commercial engagement — code and architecture stay private; details available on request.

02
Open-core Design partners Kubernetes
EvalCI — "pytest for LLM quality"
CI-native evaluation and regression-detection platform for LLM applications.
+
EvalCI dashboard screenshot
Overview

Instead of comparing scores over time and hoping drift is visible, EvalCI tests against an explicit baseline and fails a build the way a broken unit test would. The detection engine, evalci-core, is open-core, Apache-2.0 licensed and pip-installable, self-hosted from day one via Docker Compose or Kubernetes. It's positioned directly against Braintrust, LangSmith, and Langfuse, and is currently in early access with design partners.

Key contributions
  • Built the platform from scratch across five phases: FastAPI, Celery, PostgreSQL, Redis, Kubernetes, and a Next.js dashboard
  • Migrated the scheduler from APScheduler to Celery Beat, uncovering and fixing several deep infrastructure bugs
  • Open-sourced the regression-detection engine with a 28-item golden dataset spanning four domains
  • Built billing (Paystack and Flutterwave), a GitHub Action for automated PR comments, and a 15-minute self-host quickstart
Stack
PythonFastAPIKubernetes CeleryNext.js

pip install evalci-core

03
Deployed Confidential Next.js
ChuksForge FinOps — AI-Native Financial Operating System
Double-entry ledger and billing infrastructure built for how AI-heavy companies actually spend.
+
FinOps ledger screenshot
Overview

A real double-entry ledger with an 80-account Nigerian-compliant chart of accounts, six built-in billing adapters (OpenAI, Anthropic, Vercel, Railway, Supabase, OpenRouter), and Paystack billing on top. The kind of financial infrastructure most AI startups improvise in a spreadsheet until it breaks.

Key contributions
  • Designed the double-entry ledger and chart of accounts from scratch, plus a priority-queue rules engine and RBAC
  • Built six billing adapters for the platforms AI companies actually pay for
  • Deployed to Railway with Supabase, resolving PgBouncer prepared-statement errors, a JWT persistence bug, and orphaned queue-connection issues
  • Added manual accounting tools: a cost-event modal, a Nigerian-bank-format CSV import wizard, and 15 journal-entry templates
Stack
Next.jsExpressPostgreSQL PrismaPaystack

Confidential commercial engagement — code and architecture stay private; details available on request.

04
Portfolio build LangGraph 511 tests
Multi-Agent Data Pipeline — Portfolio Analytics Engine
A multi-agent data pipeline built to production standard.
+
Pipeline architecture diagram
Overview

A LangGraph-based multi-agent pipeline for portfolio analytics, using DuckDB and polars for fast local analytics and FastAPI to serve the results, with a real test suite behind it rather than a hand-wavy demo.

Key contributions
  • 511 passing tests at 84% coverage
  • Delivered with a full architecture diagram and developer handoff documentation
Stack
LangGraphDuckDB polarsFastAPI
View on GitHub ↗
05
Open-core LangGraph Public repo
Code Review Agent — Noise-Reduced LLM Feedback for CI
Filters and prioritizes automated code review findings to surface only actionable, high-signal issues.
+
Code Review Agent output sample
Overview

A multi-agent code review system that addresses the core failure mode of LLM-based review tools: noisy, repetitive, low-priority findings that slow development down rather than speed it up. A critic loop re-evaluates initial LLM output against static analysis results, deduplicates findings, and produces a ranked report suitable for direct CI/CD consumption.

Key contributions
  • Designed the Planner → Critic → Rewriter → Reconciler agent pipeline in LangGraph
  • Built a critic loop that re-evaluates LLM findings against deterministic static analysis before surfacing results
  • Implemented confidence-scored deduplication and severity-based prioritization
  • Applied an open-core strategy: public tooling and schemas, private prompt architecture and calibration logic
Stack
PythonLangGraph Static analysisStructured outputs
View on GitHub ↗
06
Shipped Confidential PostGIS
DispatchOS — Nigerian Last-Mile Logistics Platform
A logistics operating system for last-mile delivery: dispatch, rider management, order tracking.
+
DispatchOS dashboard screenshot
Overview

A last-mile logistics platform built as a Turborepo monorepo, combining PostgreSQL/PostGIS for geospatial dispatch logic with Mapbox for routing, Paystack for payments, and Termii for SMS.

Key contributions
  • Built the dispatch, rider-scoring, and order-tracking core across a five-phase delivery
  • Post-build inspection pass found and fixed four confirmed bugs, including a seed-data sequence collision
Stack
PostgreSQL / PostGISMapbox PaystackTermii

Confidential commercial engagement — code and architecture stay private; details available on request.

Additional work

More projects

Experience

Building in public

2024 — Present
Founder & Principal Engineer — ChuksForge AI Solutions Ltd.

Founded and lead a registered AI engineering consultancy delivering production-grade LLM systems, multi-agent architectures, and AI SaaS platforms for startups and growth-stage companies, with a focus on reliability and real-world deployment constraints.

  • Architect and deploy multi-tenant LLM systems with secure user isolation, retrieval pipelines, and evaluation frameworks
  • Design agentic infrastructure, embeddable AI components, and production-ready SaaS solutions
  • Build and operate a portfolio of AI products under the ChuksForge brand, including BOS, live with Nigerian merchants today
  • Lead production hardening and scale-readiness audits for live systems, diagnosing infrastructure bottlenecks before they reach users
Ongoing
Systematic Trader (Independent) — Forex & Crypto

Apply rule-based and discretionary strategies in high-noise, probabilistic environments across forex and crypto markets.

  • Design and test structured trading systems under uncertainty, optimizing for risk-adjusted outcomes
  • Build custom Pine Script strategies and AI-assisted market analysis workflows
  • Carry the same discipline: signal vs. noise, probabilistic thinking, and risk control, directly into AI system design
2025
Prompt Engineering Research — 9-Project Lab

A structured research initiative into LLM behavior, reliability, and system integration patterns, later folded into the applied projects above.

  • Built a multi-project prompt engineering lab covering summarization, style transfer, RAG, and hallucination mitigation
  • Developed reusable evaluation patterns, prompt frameworks, and shared utilities
Get in touch

Ready to build something that has to work?

I'm open to consulting engagements, contract work, and select full-time AI systems engineering roles.

Email Telegram GitHub Twitter / X
Get in touch →