Local AI Transcription: Keep Audio and Transcripts Private
How to turn meetings, interviews and recordings into private transcripts, summaries and a searchable archive without sending the audio to a public API.
Senior backend engineer · PHP & Go · Europe
Senior backend engineering for PHP systems that outgrew their architecture
I work with European engineering teams on two things: deciding what to do with a legacy PHP system, and adding senior backend capacity to a team that is short one experienced engineer. Same person, same B2B agreement.
- €400/day embedded backend capacity
- €1200 fixed Migration Readiness Audit
- CET/CEST working-hours overlap
- EUR invoice monthly, one B2B agreement
monolith/ ├── billing keep in PHP · upgrade framework ├── admin keep in PHP · low traffic ├── search extract to Go · latency + CPU ├── event-pipeline extract to Go · throughput └── reporting keep, but move off the primary DB decision first, code second
How we can work together
Engineering engagements, not staffing products. You talk to the person doing the work from the first call.
Offer A
Legacy PHP modernization & PHP → Go migration
I help engineering teams decide what should stay in PHP, what should be modernized, and what is worth extracting into Go. The audit comes before the code, and the honest answer is often "modernize this, move only that".
- Migration Readiness Audit, €1200, fixed price
- PHP → Go implementation, from €550/day
- PHP modernization work, €400/day
- Fractional migration lead, from €550/day
See how the audit works
Offer B
Senior backend engineer inside your team
Senior backend capacity without opening another payroll position. PHP, Go, APIs and distributed backend systems, working in your repository, your board and your review process.
- €400/day, up to 5 days per week
- 10-day paid pilot to start
- One B2B agreement, monthly EUR invoice
- CET/CEST overlap, English in writing and review
See how the engagement works
Offer C
Private AI and local LLMs
Useful AI workflows on infrastructure you control: transcription, document search, summarization and internal assistants. Local when privacy or economics justify it; cloud when it makes more sense.
- Private AI Readiness Audit, from €1500
- Single-workflow pilots, from €3500
- Production deployment, from €650/day
- Local, on-premise, private cloud or hybrid
See private AI deployment
Position
Your PHP monolith does not necessarily need a rewrite
A rewrite is a decision with a multi-year cost, and it is the wrong default for most systems. I am paid to make the right architecture call, not the largest one.
01 · Measure
Find the actual constraint
Latency, throughput, database contention, deployment friction, or simply the cost of changing the code. Each of those has a different fix, and only some of them involve Go.
02 · Separate
Split the codebase by consequence
Most of a monolith is fine where it is. The interesting part is the handful of components whose load profile no longer matches the runtime they run on.
03 · Decide
Write the decision down
Keep, modernize, or extract, per component, with the reasoning, the risk and the sequencing. That document is what your team and your board can argue with.
What I work with
Backend only. No design, no frontend frameworks, no promises about parts of the system I do not touch.
Working stack
- PHP
- Symfony
- Laravel
- Go
- PostgreSQL
- MySQL
- Redis
- REST / gRPC APIs
- RabbitMQ
- Docker
- Kubernetes
- CI/CD
Remote from Serbia, working hours overlapping CET/CEST. Engagements run on one B2B agreement with a monthly invoice in EUR.
How an engagement starts
01
Call
30 minutes on the system, the constraint and the deadline. No sales layer in between.
02
Scope
A written scope: audit, implementation, or embedded capacity, with the price and what is out of scope.
03
Agreement
One B2B service agreement. NDA on request. Invoiced monthly in EUR, no intermediaries.
04
Start
Repository access, one real ticket, and a first pull request inside the first days of the engagement.
Blog
How I think about this work
Local LLMs for Business: When Running AI Yourself Makes Sense
Running a local model is easy. Running one that reliably improves a real workflow is not. This is the decision framework before anyone buys a GPU.
Local LLM Hardware Requirements: What Do You Actually Need?
How model size, quantization, VRAM, RAM and concurrency interact, with workload profiles you can update as hardware and models change.
Next step
Tell me what the system is doing
Stack, the constraint you are hitting, and the deadline you are working against. If a rewrite is the wrong answer, you will hear that on the first call.