Chibuike Dialaekwe
AI Automation Intern | Erste Hausverwaltung GmbH, Köln
I build AI agents and n8n workflows that take repetitive work off people's desks.
Workflows I Build Myself
My n8n Nebenkostenassistent reads Google Sheets, creates one PDF per tenant, emails it and logs every send. It is my own hands on work, and you can see the workflow here.
Self Initiated Projects
Reads Google Sheets, creates one PDF per tenant, emails it and logs every send.
Open Project →Uses ScraperAPI, Python and JavaScript code nodes, and AI ranking.
XGBoost forecast plus an AI agent. Cut a simulated planning task from 4 hours to 45 minutes.
Open Project →Evaluates Claude API classifiers with LangSmith test datasets.
Live Projects
20 live projects you can open in the browser. The web tools were built with Claude Code and Codex: I set the problem and the AI wrote the code. The n8n workflows are my own hands on work. Most of the tools tackle logistics and inventory problems, which I know from my warehouse work.
Data and scenarios use public datasets or altered figures for demonstration. All tools are fully functional.
Projects on this site are built on public or simulated datasets to demonstrate methodology. Metric figures are illustrative of tool capability, not outcomes delivered at a company. Happy to walk through the logic of any project.
Property managers spend hours auditing utility bills, calculating proportional tenant splits based on area or headcounts, compiling statements, and emailing them manually.
▶ SOLUTIONn8n workflow triggered monthly. It reads data from Google Sheets, loops through tenants, drafts HTML statements, calls the PDFShift API to generate PDFs, emails them via Gmail, and logs every send.
One PDF per tenant, emailed and logged
Demand planners spend 70% of their time on routine replenishment decisions reviewing stock levels and creating POs instead of strategic planning.
▶ SOLUTIONXGBoost forecast plus an AI agent that turns inventory levels into ranked purchase recommendations.
Cut a simulated planning task from 4 hours to 45 minutes
Traditional monthly forecasting misses sudden demand signals from market events, competitor promotions, or weather patterns.
▶ SOLUTIONBrowser demo comparing a statistical forecast with a demand sensing forecast, with anomaly alerts on sample data.
detects demand shifts 3 weeks early 38% MAPE improvement over statistical baseline on sample data
Port congestion and transit delays cause reactive, costly disruptions with no early warning system. Companies only discover delays after goods are already stuck.
▶ SOLUTIONBrowser dashboard scoring each shipment lane by risk level, flagging emerging risks 48 to 72 hours before they escalate.
48 to 72 hours early risk detection Supports proactive rerouting across multiple shipment lanes
Seasonal demand spikes are consistently underplanned, leading to lost sales in peak months and overstock in off peak months. Finance and supply chain use different numbers.
▶ SOLUTIONBrowser demo of a seasonal demand forecast with 12 month estimates, confidence intervals and a scenario toggle.
22% MAPE improvement on sample data 12 month forward planning across simulated product categories
Manual reorder point calculation leads to delayed purchase orders and frequent stockouts. Static reorder points don't adjust to changing demand or lead time.
▶ SOLUTIONBrowser tool with dynamic reorder point formulas, colour coded alert list, and draft PO generation using Z score safety stock logic.
full elimination of manual reorder tracking 100% automated reorder triggers for 200+ SKUs
Retailers face costly stockouts and overstock due to volatile demand and manual replenishment. No ABC/XYZ segmentation all SKUs treated equally.
▶ SOLUTIONBrowser dashboard with replenishment alerts driven by ABC/XYZ analysis and safety stock calculation across 500+ sample SKUs.
up to 40% reduction in stockout events (on sample data) Handles safety stock sizing for 500+ SKUs
Procurement teams lack visibility into supplier pricing trends, leaving cost saving opportunities unidentified. Spend data is fragmented across multiple systems.
▶ SOLUTIONBrowser tool with sample PO data, supplier comparison tables, and alerts when prices exceed category average by >10%.
spend analysis turnaround from 3 days to 2 hours Surfaces savings opportunities (~€0.8M on sample data)
Last mile delivery accounts for 53% of logistics costs, yet route planning is done manually. Drivers take suboptimal routes, causing excess fuel consumption and missed windows.
▶ SOLUTIONBrowser routing demo using Berlin district geodata. Calculates optimal delivery sequences for 5 vans across 50 daily stops.
18% delivery cost reduction on demo data Route consolidation potential across 5 city zones
Companies discover supplier issues only after delays occur. No standardised scoring framework exists for proactive risk management.
▶ SOLUTIONBrowser scorecard rating 30 sample suppliers across 5 dimensions with composite risk scores and alerts below threshold.
early risk warnings for ~12% of a supplier base before disruptions escalate
Warehouse managers lack a unified view of pick rates, accuracy, and throughput across shifts. Data requires manual Excel manipulation to report.
▶ SOLUTIONBrowser dashboard showing 12 warehouse KPIs with shift level filters on sample data.
daily real time OTIF tracking Pick accuracy improvements from 94% to 97.3% modeled on sample data
S&OP is done in disconnected spreadsheets, causing misalignment between forecasts and production capacity. No shared data model exists.
▶ SOLUTIONInteractive browser S&OP simulator reconciling demand forecasts with production capacity constraints and inventory positions.
up to 30% improvement in supply demand alignment Consensus planning structure across product families
Freight costs are fragmented across multiple carriers and invoice systems, making total cost of shipment nearly impossible to benchmark.
▶ SOLUTIONBrowser tool comparing freight invoices from 8 sample carriers, benchmarking each lane and flagging cost more than 15% above benchmark.
surfaces hidden freight cost (~€180K on sample data) Detects carrier overpayment patterns across multiple lanes
Purchasing teams have no historical visibility into supplier lead time accuracy. Promised lead times in contracts often differ from actual performance.
▶ SOLUTIONBrowser tracker comparing promised vs actual lead time per supplier, calculating reliability scores and flagging >20% variance for review.
up to 25% on time delivery improvement Identifies top chronic late suppliers for performance management
EU CSRD regulations require Scope 3 emissions reporting, but emissions data is fragmented across carriers and not calculated consistently.
▶ SOLUTIONBrowser tool calculating per shipment CO₂e using GLEC Framework. Includes modal shift scenario comparison (road vs rail vs air).
Scope 3 emissions quantification across 8 logistics lanes ESG ready reporting structure
Companies manage inventory metrics in isolation with no visibility into whether turnover ratios are competitive versus industry benchmarks.
▶ SOLUTIONBrowser tool comparing inventory turnover, DIO, dead stock %, and GMROI against sector benchmarks with RAG status.
surfaces trapped working capital (~€2.1M on sample data) ITR benchmarking across 4 industry sectors
Multi depot distribution networks are rarely re evaluated. Depot locations established years ago may now be generating excess transport cost.
▶ SOLUTIONBrowser tool using a weighted centre of gravity method to compare the current network with 3 scenarios: 1, 2, or 3 depots.
18% distribution cost reduction on sample data DC closure scenario with €1.4M/yr savings on sample data
Food and pharma companies lose millions to write offs due to poor batch visibility. FEFO picking is not consistently enforced across locations.
▶ SOLUTIONBrowser tool showing batch expiry data, risk tiers and recommended redistribution or markdown actions on sample data.
prevents product write offs (~€380K on sample data) FEFO compliance improvement up to 55%
Companies invest in demand planning but never measure forecast accuracy systematically. Planners repeat the same systematic biases each cycle.
▶ SOLUTIONBrowser scorecard tracking MAPE, forecast bias, and tracking signal by product family, planner, and customer on sample data.
up to 30% forecast error reduction Planner level bias decomposition across planning teams
Purchase orders are tracked in ERP but data is never surfaced into actionable analytics. PO cycle time, compliance, and spend concentration are invisible.
▶ SOLUTIONBrowser hub showing PO cycle time, spend by supplier, compliance rate, and open PO aging on sample data.
real time tracking of 1,200+ POs 41% reduction in 3 way match exceptions on sample data
About Me
I am an AI Automation Intern at Erste Hausverwaltung GmbH in Köln. In my own projects I look for work people repeat by hand and build something that does it for them, mostly with n8n.
I did not start in tech. I ran operations at a lounge in Nigeria, then worked in warehouses, logistics and events in Berlin while studying for my Master's. That background helps me ask the right question before I build anything: who does this work today, and what goes wrong when they are busy?
I also test before I trust. I wrote a Python script that evaluates Claude API classifiers against LangSmith test datasets. I am looking for a full time junior role in AI automation, workflow automation or software integration.
Work ExperienceOfficial Work References / Arbeitszeugnisse
Let's Talk
I am looking for full time junior roles in AI automation, workflow automation, AI engineering, software integration and low code. I am based in Köln. Berlin and NRW are my preferred locations, and remote is welcome.