About me
I’m a naturally curious person so I keep trying to find better ways of doing things.
From Concepts to Reality
Many of my projects start as small ideas, which I then refine through dogfooding to test their applicability and later on with colleagues. This approach has helped me direct change and improve numerous processes.
Process Improvements
A few examples:
- From manual to automatic testing:
- Before: 6 months from hardware development through purchasing, assembling, testing the tester and executing paper and pen testing
- After: overnight testing with automatic report
- From manual to automatic timesheets checking:
- Before: accounting would identify an issue and signal to PMs that would ask the team for corrections and then back and forth until all had been solved
- After: daily synchronisation of tasks between ticketing systems and timesheet systems with automatic corrections
- From reporting day to seamless reporting:
- Before: PMs would have a day of the week destined to reporting on all tickets worked on during the last 7 days
- After: AI-generated and human-curated reports with no tickets being missed
AI Expertise
When it comes to AI, I’ve developed expertise in the following areas using Python:
- Ollama
- Multiple models running locally including multimodal and embeddings
- More recently llama3:70b, qwen2:72b and llava:34b running locally
- OCR
- To increase robustness when adding context to prompts I have added OCR using macOS built in functions
- Fast Whisper and Melo
- To facilitate interaction with LLM generated content I have used fast-whisper for transcription and Melo for Text-to-Speech
- Streamlit
- Interacting with Python scripts can get a bit convoluted for more complex actions, to mitigate that I have used streamlit for UI development
- Langchain
- Provides wrappers that can simplify the implementation of agentic solutions
- My own curated prompts
- With some trial and error I have developed my own curated prompts for my most common tasks
Exploring New Frontiers
I’ve also experimented with various AI projects, each with its own unique merits:
Privacy-Focused Approach
All my projects are designed to respect user privacy, collecting no data and running locally.
Get in Touch
You can find me on:
Keep learning!