Full-Stack & AI Dev | TypeScript, Next.js, React, Node
- $50k+
- Earned
- 5x
- Hired
- 5.0
- Rating
- 136
- Followers
Full-Stack & AI Dev | TypeScript, Next.js, React, Node
Growth Driven Design & Dev for Startups and SMBs
AI Engineer | LLM Agents - RAG & Evals - Full Stack
- 5.0
- Rating
- 20
- Followers
AI Engineer | LLM Agents - RAG & Evals - Full Stack
Principal AI Engineer | GenAI, Python, RAG & AI Agents
Senior Full-Stack & WordPress Developer · EN/FR · Remote
Engineering Dreams Into Reality - Full Stack Developer
Full-Stack Developer · UI/UX & Landing Pages
Full-Stack Developer · UI/UX & Landing Pages
Sagini AI Analytics — AI-powered data intelligence platform (sagini.net (http://sagini.net))
Designed and built the full marketing site for Sagini AI Analytics, a tool that turns any uploaded file into instant charts, insights and anomaly reports.
The visual identity here is deliberately different from a typical SaaS landing page. Instead of the common blue and sans-serif "tech" look, the site pairs a deep navy background with a warm gold accent and a serif display typeface for headlines. This gives the product a premium, editorial feel that signals trust and authority, fitting for a tool people use to make decisions on their actual data. Body copy stays in a clean sans-serif for contrast and readability, so the serif is reserved for moments that need weight.
The hero headline uses two different type treatments in one sentence: "Your data," in a solid serif, "decoded instantly" in a gold italic script. This creates a clear visual hierarchy and makes the value proposition the most memorable element on the page, rather than burying it in a paragraph. A trust bar directly below (10+ formats, under 5 seconds, AI powered, 100% your data) answers objections before the visitor has to ask.
The "How it works" section uses a three-step numbered layout with a connecting line between steps, which reads as a literal pipeline rather than a bullet list, reinforcing the "file in, insight out" simplicity of the product. The features grid below groups capabilities into six equal cards with consistent icon and copy structure, making a technically dense feature set easy to scan.
The supported-formats section replaces a wall of text with format badges (CSV, PDF, XLSX, PNG, and others), which communicates breadth instantly and visually rather than through a sentence a visitor would skim past.
The page closes with a repeated, high-contrast CTA block on the gold-accented navy background, plus an honest "Currently in Beta" and "Free during beta" disclosure, which builds trust rather than undermining it.
Stack: FastAPI, Python, Claude API. Deployed with nginx, systemd, Let's Encrypt SSL, MySQL.
2
270
Ranksit — AI content and SEO SaaS (ranksit.net (http://ranksit.net))
Designed and built the full marketing site for Ranksit, an AI tool that turns a keyword into a researched, SEO-scored article.
The hero leads with a clear before/after promise ("Turn a keyword into publish-ready, optimised content"). The key differentiator, "optimised," is pulled out in a contrasting colour so it reads first. A single input field sits directly under the headline instead of a generic CTA button, letting a visitor try the core action immediately, with popular example prompts underneath to lower the effort of that first interaction. Four short feature cards below establish credibility (researched, audited, live feedback, ready to ship) without needing a separate scroll section.
The "How it works" section breaks the product into three numbered steps for scannability, followed by real content score examples (92, 88, 74) that make the abstract "SEO scoring" concept concrete and visual rather than just a claim.
Pricing uses a simple two-tier layout with the Pro plan visually distinguished by a border and badge. The price is the largest element on the card since it's the first thing a comparison shopper looks for.
The FAQ section closes objections before the final CTA. The page ends on a repeated, action-oriented close ("Ready to rank?") rather than trailing off after pricing.
Stack: FastAPI, Python, HTMX.
1
233
GM Rides — Transport booking platform (Kenya)
Designed and built the marketing site and booking flow for GM Rides, a platform for intercity buses, airport transfers and private hire across Kenya.
The hero establishes trust before it establishes the product. A status badge ("Kenya's Most Modern Transport Platform") sits above a large two line headline, with the differentiator word ("Smarter") set in a lighter blue tone against a dark navy background for contrast. The supporting line answers the three questions a first time visitor has immediately: what can I book, where does it work, and how do I pay.
The booking widget is placed directly in the hero rather than behind a separate page, with tabs for One Way, Round Trip and Airport Transfer so the most common use case is one click away. Trust stats (towns covered, happy travellers, on time rate, support hours) sit just below the fold, using large numbers with short labels so they scan in under a second.
Further down, the "Travel in Comfort" and "Built Different" sections use a consistent card pattern, an icon, a short title, and one line of supporting copy, repeated across six feature cards. This keeps the page easy to skim rather than forcing the visitor to read paragraphs, and each card answers a specific objection a traveller might have (safety, comfort, punctuality, payment flexibility).
The footer closes with practical reassurance: multiple accepted payment methods shown as logos, direct contact details, and full service/site navigation, so a visitor who scrolls all the way down still has a clear next step.
Stack: Next.js, Laravel, MySQL. Payments: M-Pesa STK Push, Stripe. Multi currency support (KES, GBP, USD, EUR).
2
232
AI Systems for Agents, Voice, Retrieval & Automation
AI Operations Copilot | RAG, Agents & Workflow Automation
Role:
AI Systems Engineer | LLMs, RAG, Backend & Automation Architecture
An AI operations copilot that connected CRM data, Notion, internal knowledge, and third-party APIs into one context-aware operational system.
I designed a RAG pipeline with agent orchestration to retrieve business context, reason across workflows, generate structured recommendations, and execute approved actions through connected tools.
The system supported natural-language queries, automated reporting, workflow recommendations, and operational task execution.
A FastAPI backend handled asynchronous processing, API coordination, structured logging, retry logic, and failure recovery.
The result was a unified operational layer that reduced repetitive manual work and helped teams retrieve information, make decisions, and trigger actions more efficiently.
0
54
AI Calling Agent | Real-Time Voice AI & Human Escalation
Role:
AI Systems Engineer | Voice AI, RAG & Backend Infrastructure
A real-time AI calling agent for customer support, lead qualification, and automated phone conversations.
The system combined live speech recognition, LLM reasoning, RAG-based knowledge retrieval, natural voice synthesis, and conversation-state management inside a low-latency asynchronous pipeline.
I designed the workflow to understand callers, retrieve approved business context, collect structured information, respond naturally, and escalate complex conversations to a human when needed.
The architecture included interruption handling, fallback logic, caching, transcript logging, observability, and resilient API orchestration.
The result was a controlled voice AI system that could manage routine conversations while preserving reliable human handoff for cases outside the agent’s scope.
0
68
Intelligent Invoicing | Document AI, SAP HANA & Financial Automation
Role:
Senior AI Systems Engineer | Document Intelligence, Financial Automation & SAP Integration
An intelligent invoicing platform that automated accounts payable and receivable workflows while integrating directly with SAP HANA.
The system used OCR and document intelligence to extract invoice data, validate fields, identify duplicate or suspicious submissions, and generate AI-assisted GL coding recommendations.
Invoices were routed through configurable approval workflows based on business rules, user roles, invoice value, and exception conditions.
The platform also included supplier performance scoring, SLA monitoring, cash-flow forecasting, role-based access, and complete audit trails designed to support financial governance and SOX-404 compliance.
The result was a more controlled financial workflow with reduced manual handling, stronger traceability, and clearer exception management.
0
46
AI Voice Agent | Inbound, Outbound & CRM Automation
Role:
AI Systems Engineer | Voice AI, Backend Architecture & CRM Automation
AI voice agent system for automating inbound and outbound business calls across customer support, appointment booking, lead qualification, and sales workflows.
The system processed live audio through low-latency speech-to-text, used LLM reasoning for context-aware conversations, and generated natural spoken responses through text-to-speech.
It maintained conversation state, handled interruptions and fallback scenarios, and captured structured information during each call.
I connected the voice workflow to CRM systems and external APIs so the agent could update records, qualify leads, schedule appointments, and trigger follow-up actions.
The result was a reusable voice automation system that connected customer conversations directly with business operations.
0
49