Four projects, each with the same shape: a real problem, a deliberate approach, the tools I reached for, and a result that mattered. See the role-by-role breakdown →
2025 — Present · Open-sourceCreator & maintainer
Draftly — an AI writing assistant that sounds like you
A Chrome extension (Manifest V3) that humanizes, replies, translates and fixes grammar right inside Gmail, LinkedIn and WhatsApp, with a bring-your-own-key model for Claude, GPT, Gemini and more.
Stack · JavaScript · LLM APIs · Chrome MV3 · UX
The challenge
AI writing tools are powerful but locked in separate tabs, and most send your text to a vendor's servers. People want help where they already write, without surrendering privacy.
What I built
A single floating panel that follows you across the web. Pick a mode (Humanize, Reply, Translate, Grammar), click once, edit, apply. It reads the whole conversation thread to draft replies that fit, and you bring your own API key, so text never routes through Draftly's servers.
A standardised sales pipeline and reporting system for a global distribution network spanning B2B distributors and B2C residential customers.
Tools · Odoo CRM · Excel · Google Sheets · Meta Ads
The challenge
Across 60+ countries and a 10-person sales team, deal stages meant different things to different people, so leadership couldn't tell where revenue actually stalled.
The approach
I architected a 16-stage Odoo CRM pipeline with crisp stage definitions, layered a three-tier vendor classification on top, added geo-linguistic lead routing, and built Month-on-Month performance dashboards with per-rep drilldowns.
Odoo CRMFunnel designExcelSOPs
The result
One clear, shared view of the funnel, and faster, fairer coaching and quota planning.
A unified analytics layer and automated data-quality pipeline supporting 200+ marketing campaigns.
Tools · Python · SQL · ETL · Google Analytics
The challenge
Five disconnected sources (CRM, website analytics and more) never reconciled, so reporting was slow and decisions were built on shaky numbers.
The approach
I integrated all five into a single reporting layer and automated missing-value handling across platforms, then rebuilt traffic and behaviour reporting for speed.
PythonSQLETLAutomation
The result
Trusted data, delivered fast, and a marketing engine that could finally adapt in real time.
▲ +40% data accuracy · −60% report time · −20 hrs/mo manual work
2022–2025 · Nexford UniversityLondon, UK · Remote
Power BI dashboards that grew sales 46%
A decision-support system for sales and admissions that aligned ten teams around one trusted view of the funnel.
Tools · Power BI · SQL · Excel
The challenge
Sales and admissions were flat, and ten teams each read the data their own way, so strategy stalled in debate.
The approach
I built and maintained Power BI dashboards of the key metrics, evaluated each team's data needs, ran competitor analysis, and added a milestone-tracking tool, all on a foundation of 100%-accurate reporting.
Power BISQLKPI designForecasting
The result
Clear, shared decisions that moved the top line and the market position.