# Raynan Wuyep > Raynan Wuyep is an AI engineer who builds agents that do real work, and the infrastructure and systems that power them. As a founding engineer at DramaLabs he took Nisa (an AI co-worker that lives in Slack, WhatsApp and iMessage) from prototype to production: the agent harness, company-brain memory, approval-first autonomy, channels, evals and QA. He builds open-source agents that operate real computers (Clacky on Windows, cell-use on iPhone, shoal in the browser) and founded Outbid and ATS-Zero, used by 5,000+ people across 40+ countries. Based in Abuja, Nigeria; works remotely. Contact: me@raynanwuyep.com. Résumé: https://raynanwuyep.com/Raynan_Wuyep_Resume_2026.pdf. Location: Abuja, Nigeria (remote). ## Current roles - Founding Engineer, DramaLabs (Jul 2026 – present, Germany, EU (remote)): Took Nisa AI from prototype to production: agent harness, Slack / WhatsApp / iMessage channels, company-brain memory, approval-first autonomy, evals + QA. - Founding Product Engineer, Stealth AI startup (Oct 2025 – present, San Francisco (remote)): Production multi-agent AI platform, owned end-to-end from prototype to production (specifics confidential). ## Flagship work - [Nisa AI · DramaLabs](https://dramalabs.co): The AI co-worker that texts you first. Founding Engineer · DramaLabs, Jul 2026 – present. Nisa is an AI chief-of-staff for small teams that lives where they already talk: Slack, WhatsApp and iMessage. She reads an authored company brain, runs scheduled and event-triggered work across the team's apps, and hands back drafts and decisions, never sending anything without an explicit approval. Raynan took Nisa from prototype to production-grade himself: the agent harness, the channels, the memory, the approval layer and the infrastructure around them. How it works: a message from Slack, WhatsApp or iMessage enters one agent loop. Nisa pulls context from the company brain, picks the right skill, calls the team's tools over MCP (sandboxed when it runs code), and answers on the same channel. Anything that would leave the building comes back as an approval card first, and the turn is distilled back into memory. She's proactive by design: scheduled skills send the daily brief, flag stalled threads and deals, and prep meetings before anyone asks. That's the "texts you first" part. Highlights: Agent harness: multi-model tool-calling loop with provider failover; script skills run in sandboxed microVMs, agent skills as tool-calling loops; Skills: 40 of them (SKILL.md + references + scripts), run on a schedule, on a trigger, or mid-chat via /name, plus one that writes new skills from a plain-language request; Company brain: versioned markdown per team and per person, typed facts with provenance, hybrid keyword + vector retrieval; Approval-first autonomy: every outbound action is an Approve / Edit / Cancel card, one code path gates every surface, and standing "always allow" permissions are narrowly scoped; Self-maintaining memory: every conversation turn is distilled into memory; a nightly "dream" pass retires stale facts, and nothing merges without approval; Channel-native: Slack Block Kit cards with live status and a stop button, WhatsApp interactive buttons, iMessage approvals, voice notes on every channel; Tools & isolation: 3,000+ tools via MCP and Composio, behind multi-tenant, fail-closed isolation; Observability & QA: Langfuse tracing, golden-set and LLM-judge evals; production deploys are blocked until a cross-channel QA simulation passes on the exact commit. Stack: Python, FastAPI, Agents SDK, LiteLLM, MCP, Postgres, Langfuse, Next.js. - [Multi-Agent AI Platform](https://raynanwuyep.com/#project-agent-platform): A production agent swarm at a stealth SF AI startup. Founding Product Engineer · stealth SF AI startup, Oct 2025 – present. Founding engineer on a production multi-agent AI platform, shipped from prototype to live. A LangGraph supervisor coordinates a swarm of specialised agents end-to-end, backed by rate-limited background processing and full production observability. The company and product specifics are confidential. Highlights: Orchestration: a multi-agent swarm under a LangGraph supervisor with persistent state; Background processing: rate-limited Celery + Redis queues with fork-safe async workers; Integrations: multiple third-party APIs with backoff + retries; Full observability: LangSmith · Sentry · OpenTelemetry · Prometheus. Stack: Python, FastAPI, LangGraph, Celery, Redis, PostgreSQL. - [ATS-Zero](https://atszero.com): An AI career copilot that applies in 15 seconds. Founder & Engineer, 2025 – present. A Manifest V3 Chrome extension that reads any job posting and generates an ATS-optimised, tailored resume, turning a 15-minute application into 15 seconds. Heavy AI work runs serverless to keep the client footprint tiny. Highlights: 2,000+ active users; 50K+ views on the launch demo; 15-minute apply flow → 15 seconds; Two-stage DOM engine; React UI isolated via Shadow DOM; Serverless AI pipeline on Supabase Edge Functions. Stack: TypeScript, React (Plasmo), Supabase, Chrome MV3, LLM APIs. - [Outbid](https://useoutbid.com): Real-time Upwork job intelligence for freelancers. Founder & Engineer, 2026 – present. A SaaS that alerts freelancers to new Upwork postings within 60 seconds and drafts AI-tailored proposals so they apply before the competition. Full production stack on AWS, dual-currency billing, and a resilient real-time ingestion pipeline. Highlights: 3,000+ users across 40+ countries; 11.7M+ alerts from 312K+ jobs scanned; 5,800+ jobs/day processed; alerts within 60s; Dual billing: Stripe (USD) + Paystack (NGN). Stack: Python, asyncio, FastAPI, PostgreSQL, Docker, AWS EC2, Telegram, Stripe. ## Open source and launch demos - [Clacky](https://raynanwuyep.com/clacky): An AI buddy on your Windows PC that sees, points, and does. Creator & Maintainer. The open-source Clicky for Windows, built on Claude. Hold a hotkey and talk: she answers about whatever is on your screen, points at the exact UI element, and carries out multi-step tasks with computer use. Teach her a routine once by voice and she keeps it as a skill. Highlights: 66★ on GitHub; 63K+ views on the launch demo; Pixel-accurate pointing that snaps to real UI elements via Windows UI Automation; Voice-taught routines saved as SKILL.md; background agents that hand back real files; Reversible actions: "clean up my desktop", then "undo". Stack: Python, PyQt6, Claude Computer Use, Deepgram, MCP. Also: GitHub: https://github.com/Raynan00/clacky. - [cell-use](https://github.com/Raynan00/cell-use): Your AI agents can now use your iPhone. From your iPhone. Creator & Maintainer. An open-source Swift SDK for agents that see, tap, swipe and type across iPhone apps. The screenshot-and-input loop runs on the phone itself, with no computer attached; bring any model, on-device or cloud. Highlights: 20★ · fully on-device control loop behind one PhoneAgent protocol; Embedded packet-tunnel kit, so users don't install a separate VPN app. Stack: Swift 6, SwiftPM, Network Extension, Rust. Also: Docs: https://github.com/Raynan00/cell-use/tree/main/Docs. - [shoal](https://github.com/Raynan00/shoal): 1,000 AI users break your website before real ones do. Creator & Maintainer. A swarm of persona-driven AI users that test your site in real browsers, log friction as it happens and rank it into a live friction map, all rendered as a pixel-art aquarium. Multi-user scenes surface race and oversell bugs. Highlights: Up to 1,000 sessions on one laptop at ~57 MB per agent (vs. 250–500 MB norm); 9 personas × 7 strategies, accessibility persona, MCP server for Claude Code / Cursor. Stack: TypeScript, Playwright, Claude, React, WebSockets. - [LocalPDF](https://apps.microsoft.com/detail/9NL235Q73VTR): Right-click, make PDF. Nothing ever gets uploaded. Creator & Maintainer. Free offline PDF tools for Windows and Mac: merge, compress, split, rotate, convert to and from Word/JPG, unlock and password-protect, straight from the right-click menu. The app contains no networking code at all. Highlights: Published on the Microsoft Store; Windows + macOS installers; qpdf compiled in, PDFium + headless LibreOffice bundled; AES-256 protect. Stack: Rust, Tauri 2, React, qpdf, PDFium. Also: GitHub: https://github.com/Raynan00/localpdf. - [This portfolio](https://raynanwuyep.com/#project-portfolio): A portfolio you can talk to. Designer & Engineer. raynanwuyep.com: an editorial-brutalist portfolio with an opt-in agent layer. Visitors can chat with a text agent or open a real-time video avatar of Raynan that answers questions and tours his projects. Highlights: Real-time video avatar with server-minted sessions + cost guardrails; Guided tour: the agent surfaces project cards as it talks. Stack: Next.js 16, TypeScript, Anam, Gemini, Cloudflare Workers. ## More AI / agent tooling - [sandpy](https://sandpy.vercel.app): Run Python in the browser for your AI agents, for free. Creator & Maintainer. A browser-native Python sandbox built on Pyodide/WASM, letting AI agents execute Python client-side as a free alternative to E2B/Modal. Ships with LangChain and Vercel AI SDK integrations. Highlights: npm package · 1,100+ downloads in its first year; Snapshots, timeouts, pandas + matplotlib artifacts. Stack: TypeScript, Pyodide, WebAssembly, LangChain, Vercel AI SDK. Also: GitHub: https://github.com/Raynan00/sandpy, Article: https://dev.to/raynan_wuyep_14a8b980c3d9/your-ai-agent-can-run-python-for-free-heres-how-1p5d. - [LifeGraph](https://raynan00.github.io/LifeGraph/): Your photos & transactions as a behavioural map. Creator. Clusters thousands of personal photos and transactions into behavioural categories with CLIP embeddings, rendered as an interactive map. Runs locally on-GPU for privacy. Highlights: 4,381 transactions + photos clustered; CLIP · HDBSCAN · UMAP · local LLM. Stack: Python, CLIP, HDBSCAN, UMAP. Also: GitHub: https://github.com/Raynan00/LifeGraph. - [CPG Compliance Bot](https://github.com/Raynan00/CPG-compliance-bot): Parallel multi-agent compliance review for promos. Creator. Runs FTC, retailer, and brand compliance checks in parallel over CPG promotional proposals using a LangGraph multi-agent graph. Highlights: LangGraph + Claude; Parallel FTC / retailer / brand agents. Stack: Python, LangGraph, Claude. - [Meridian](https://github.com/Raynan00/Meridian): A privacy-first RAG agent for personal finance. Creator. A retrieval-augmented finance assistant built on LangChain and Google Gemini, designed to keep personal financial data private. Highlights: LangChain + Gemini RAG; Privacy-focused design. Stack: Python, LangChain, Gemini, RAG. - [pr-audit](https://github.com/Raynan00/pr-audit): AI triage that ranks PRs and plans the merge. Creator. Helps OSS maintainers triage incoming pull requests: ranking them, running an LLM review, and producing a stack-aware merge plan. Highlights: LLM review + ranking; Stack-aware merge plan. Stack: Python, LLM. ## Skills - AI agents & LLM systems: Agent harnesses & tool calling, MCP, Multi-agent orchestration (LangGraph, OpenAI Agents SDK), Agent memory & RAG (hybrid keyword + vector), Evals (golden sets, LLM-as-judge), Computer-use agents (desktop, iPhone, browser), Real-time voice agents - Languages: Python, TypeScript, Swift, Rust - Backend & infra: FastAPI, PostgreSQL, Redis, Celery, Docker, AWS, Cloudflare Workers, Sandboxed microVMs, CI/CD - Apps & frontend: React, Next.js, Tauri, PyQt6, Chrome extensions (MV3), Playwright - Observability: Langfuse, LangSmith, Sentry, OpenTelemetry, Prometheus ## Experience - Founding Engineer, DramaLabs, Jul 2026 – present (Germany, EU (remote)): Took Nisa AI from prototype to production: agent harness, Slack / WhatsApp / iMessage channels, company-brain memory, approval-first autonomy, evals + QA. - Founding Product Engineer, Stealth AI startup, Oct 2025 – present (San Francisco (remote)): Production multi-agent AI platform, owned end-to-end from prototype to production (specifics confidential). - Software Engineer, EEMS Limited, Apr 2025 – Mar 2026 (Abuja, Nigeria): Geospatial pipeline (Sentinel, GDAL, NDVI) mapping oil-contaminated land. - Front-End Lead, Techyteams, Feb 2025 – Aug 2025 (Abuja, Nigeria): +25% performance, +40% stability; introduced testing & CI. - Full-Stack Developer, Codekraftsmen, Feb 2023 – Jan 2025 (Abuja, Nigeria): REST APIs serving thousands of daily requests; auth & schema design. ## Education and credentials - B.Sc. Computer Science, Babcock University, Nigeria, 2021 – 2024 (GPA 4.21 / 5.0) - Certification: Machine Learning, Stanford University - Certification: AI Development with Python, IBM - Award: Finalist, WEMA Bank Hackaholics 4.0 (Pixel-Push) - Award: 1st Runner-Up, BUSA Fin/Tech Hackathon (RIGEL) ## Links - [Portfolio](https://raynanwuyep.com): interactive site with a talk-to-me text agent and live video avatar - [GitHub](https://github.com/Raynan00): @Raynan00 - [LinkedIn](https://www.linkedin.com/in/raynan-wuyep): in/raynan-wuyep - [X](https://x.com/WuyepRaynan): @WuyepRaynan - [DEV](https://dev.to/raynan_wuyep_14a8b980c3d9): Writing - [Résumé (PDF)](https://raynanwuyep.com/Raynan_Wuyep_Resume_2026.pdf)