Chukwunedum Davidson
AI Systems Builder from Lagos. I ship AI-native systems end-to-end — from agent architecture through deployment, evals, and iteration — using Claude Code daily and the Anthropic API. Built in environments where the network drops, the bank times out, and the user is on 3G. It taught me what production-ready actually means.
01 · core skills
└────────────────────────┘Frontend
React, Next.js, TypeScript, Tailwind CSS, Design Systems, Figma-to-code, Component Libraries
Backend
Node.js, Python, FastAPI, Express, REST APIs, SQL, PostgreSQL, MongoDB, Redis
Cloud & Infrastructure
AWS (EKS, EC2, RDS, S3, Lambda, Route 53), DigitalOcean, Cloudflare, Docker, Kubernetes, Terraform, Helm
CI/CD & Automation
GitHub Actions, CircleCI, AWS CodePipeline, Make.com, n8n, Bash/Shell scripting
AI & Agents
Claude Code (daily), Anthropic API (text + vision), OpenAI API (GPT-4o / GPT-4o-mini), agent architecture, prompt evals, Make.com, Zapier
Observability
Prometheus, Grafana, Alertmanager, Sentry, Datadog, CloudWatch, structured logging
02 · experience
└────────────────────────┘- Designed and deployed production infrastructure from scratch serving users across 13 African countries, the US, Europe, and China — 99.9% uptime since launch
- Built CI/CD pipelines with GitHub Actions — reduced deployment time from manual hours to under 10 minutes
- Integrated Claude API for regulatory policy data extraction; Cloudflare CDN for low-latency delivery across 4 continents
- Full observability stack (Sentry, Grafana) with proactive incident detection and root cause analysis
- Built internal AI and automation tools (PrepCall, an intranet portal, a Bill-of-Materials validator) that replaced manual work for sales and solutioning teams
- Supported deals across Fortinet, Palo Alto, Huawei AICC, and Cisco Webex stacks; contributed to $190K+ in pipeline. Translated technical systems into language non-technical buyers could act on
- Led 9-engineer mentorship cohort — established code review workflows, deployment standards, and incident response documentation; 100% completion rate
- Managed cloud deployments (AWS, Vercel) and CI/CD pipelines for distributed engineering teams
- Progressed from intern to lead; owned end-to-end delivery across concurrent client projects — architecture through production deployment
- Built web applications and APIs (Node.js, JavaScript, SQL); introduced code review practices and deployment discipline to an early-stage team
03 · things i've built
└────────────────────────┘Football value-betting platform with a self-calibrating Poisson model that finds edge vs bookmaker odds and learns from resolved predictions. Three-service architecture: prediction engine (FastAPI + SQLite), web API (auth + proxy), and Telegram bot — all sharing a single brain. Covers 10 leagues, 800+ team name mappings, and auto-resolves picks on a cron schedule. ~10,930 pageviews/month. Prometheus + Grafana monitoring; Cloudflare caching to reduce origin load.
Visit site →Architected and shipped a 13-service backend that ingests policy documents across African jurisdictions and produces structured intelligence for compliance teams. Claude API drives structured extraction: raw policy PDFs / web pages → normalized schema (policies, agencies, obligations, timelines) — with evals for extraction accuracy and hallucination rate before each prompt change. Full production stack end-to-end: Node.js + PostgreSQL, Docker, GitHub Actions CI/CD, Cloudflare proxy, DigitalOcean. Claude Code as daily development driver. Built internal dashboards and playbooks so the non-technical founder can operate the system independently. 2,500+ monthly active users.
Visit site →Sole developer. Took client from idea to production.
Mentorship platform for legal professionals — matching by practice area, tracking sessions and certifications, with a centralised admin dashboard. 180+ mentor-mentee matches, 450+ sessions tracked, 85% completion rate.
Visit site →Sole developer. Took client from idea to production.
Production AI agent for job-market intelligence. Scrapes Greenhouse, Lever, and BambooHR ATS platforms, scores listings against a candidate profile with GPT-4o-mini, de-duplicates, and delivers qualified matches to a Telegram channel on a schedule. Wrote evals on ranking accuracy, de-dup rate, and false-positive rate to tune prompts and filters before and after deployment — iterated agent logic based on real-world output data. Python + Flask + SQLite.
AI desktop companion adapted from Farza's open-source fork. Extended the model layer with OpenAI (GPT-4o / GPT-4.1) and Claude vision models alongside the original Claude-text stack. Screen-aware agent that pushes live screenshots to the selected vision model and renders pointer overlays from inline [POINT:x,y] tags across multiple monitors. Swift + ScreenCaptureKit on the client; Cloudflare Worker (TypeScript) as API proxy so provider keys never ship in the binary. Multi-provider TTS fallback chain: OpenAI → Speechmatics → ElevenLabs. Real-time loop: AssemblyAI websocket transcription → streaming SSE to the chat model → streaming TTS playback, all triggered by a global push-to-talk hotkey.
View on GitHub →Claude-powered presales intelligence agent built for the DataFlex sales team. Ingests prospect signals (domain, LinkedIn, public filings) and produces deal-qualified call-prep briefs. Replaced a ~2-hour manual research process per call. Built for non-technical sales users — clean UI, minimal inputs, predictable outputs.
View on GitHub →04 · writing
└────────────────────────┘- builderswhosell.com — Essays on building products and distribution in the AI era. Recent: "What Accumulates?" (moats in the AI era) and "Thou Shalt Not Make a Machine in the Likeness of a Man's Mind" (on the limits of LLM cognition).
05 · reading
└────────────────────────┘- The Lean Startup Eric Ries
- The Mom Test Rob Fitzpatrick
- The Hard Thing About Hard Things Ben Horowitz
- 12 Rules for Life Jordan B. Peterson