Brands

Four ventures, each with its own focus

Beexium is the technology and engineering company at the center, and Techietory, Uniqartia, and Galakton are the ventures around it. Each has its own audience and focus.

18+

Client organisations

6

Countries

92%

Return & referral rate

Helios Robotics · Berlin, Germany · 2022–2024

Building the ML Team & Vision System for a 6-DOF Industrial Arm

Helios came to me with a prototype robotic arm, a vision system that worked in the lab, and a hard deadline for a pilot deployment with a German automotive supplier. Over two years as their Lead ML Engineer, I built the perception stack from scratch, scaled the team, and took the system from lab bench to six production facilities. The core challenge was making the grasp pipeline robust enough to handle the full variance of real assembly-line conditions — variable lighting, worn parts, and operators who don't want to recalibrate.

94%

Grasp success rate

6

Facilities deployed

8×

ML team growth

Full case study

Pulse Autonomy · Munich, Germany · 2020–2022

L4 Perception Stack: Sensor Fusion Under Adverse Weather

Pulse's autonomous vehicle platform was performing well on sunny test tracks and struggling in rain, fog, and low-light urban conditions. My focus was multi-modal sensor fusion — combining camera, LiDAR, and radar outputs with a learned fusion model trained on domain-randomised simulation data and fine-tuned on real-world edge cases. The key insight was treating sensor degradation itself as a learned signal, not a failure mode to mask out.

28ms

Inference latency

+41%

Rainy-weather mAP

3×

Faster retraining cycle

Full case study

Vektor Energy · Ankara, Turkey · 2025

Demand Forecasting for Grid-Scale Energy Management

Vektor operates energy distribution infrastructure across five regions of Turkey. Their existing forecasting system was a rule-based model that couldn't adapt to renewable energy variability. I designed and deployed a Temporal Fusion Transformer-based forecasting pipeline that accounts for weather, seasonal patterns, industrial demand cycles, and inter-regional dependencies — all while remaining interpretable enough for grid operators to trust and act on its outputs.

18%

MAPE reduction

5

Regions covered

Real-time

72-hour horizon

Full case study

Vektor Energy · Ankara, Turkey · 2025

Demand Forecasting for Grid-Scale Energy Management

Vektor operates energy distribution infrastructure across five regions of Turkey. Their existing forecasting system was a rule-based model that couldn't adapt to renewable energy variability. I designed and deployed a Temporal Fusion Transformer-based forecasting pipeline that accounts for weather, seasonal patterns, industrial demand cycles, and inter-regional dependencies — all while remaining interpretable enough for grid operators to trust and act on its outputs.

18%

MAPE reduction

5

Regions covered

Real-time

72-hour horizon

Full case study

ML Project Engagement

Scoped, fixed-duration projects to build and deploy a specific machine learning system, from problem framing to production handoff.

4 to 12 weeks per engagement
Full lifecycle ownership
Evaluation and handoff documentation
Four weeks of post-launch support

Embedded Engineering

I join your team for an extended period as a senior engineer, attending standups, writing code alongside your engineers, and contributing to architecture decisions.

3 to 12 month retainer
Embedded in your engineering team
Your team owns the code
Mentoring for junior engineers

Research Collaboration

Joint research projects with academic institutions, research labs, or innovation-focused companies, typically producing papers, open-source tools, or a funded prototype.

Co-authorship on publications
Grant writing support
Experimental design and analysis
Open-source deliverables

Advisory & Audit

A short, focused engagement to evaluate your current ML system or technical approach, useful before a major investment, a fundraise, or a production failure review.

1 to 2 week turnaround
Written technical report
90-minute strategy session
No obligation to continue

Interested in working together?

I have limited availability each quarter. If you have a project worth talking about, reach out and let's find out if I'm the right fit.

Start a conversation See more work

Four ventures, each with its own focus

What I've built and where it lives

Building the ML Team & Vision System for a 6-DOF Industrial Arm

L4 Perception Stack: Sensor Fusion Under Adverse Weather

Demand Forecasting for Grid-Scale Energy Management

Demand Forecasting for Grid-Scale Energy Management

Four ways to work together

ML Project Engagement

Embedded Engineering

Research Collaboration

Advisory & Audit

Interested in working together?