Symfa Team

Happy Software Tester’s Day! 🐞

Attention to detail, curiosity, patience, and persistence are just some of the qualities our QA engineers bring to our projects every day. Their ability to spot even the smallest issues makes a real difference to the quality of the solutions we deliver to our clients.

Thank you to all our QA colleagues for the care and dedication you put into your work and for helping our teams deliver high-quality solutions every day! 💚

Wishing you fewer critical bugs, more successful releases, clear requirements, and smooth working days! ☺️

Happy Software Tester’s Day!🎉

Join Symfa as a Senior DevOps Engineer!

We have a new opportunity for an experienced Senior DevOps Engineer to join a long-term project for our client in the insurance domain 🧑‍💻

One of the key focuses of the role is on-premises to Azure migration, so hands-on experience migrating .NET applications and MS SQL databases to Azure is especially important.

What we’re looking for:

☁️ 4+ years of experience in DevOps / Infrastructure Engineering
☁️ Strong hands-on experience with Microsoft Azure and cloud infrastructure
☁️ Practical experience with on-prem → Azure migrations
☁️ Experience with Azure DevOps Pipelines or similar CI/CD tools
☁️ Knowledge of networking, security, RBAC, and IaC
☁️ English at B2+ level

Interested? Send your CV to [email protected] 📩

Know someone who might be a good fit? Please share this post with them 🙌

At Symfa, our colleagues live and work in 20+ countries around the world. Each place comes with its own little surprises, and sometimes even cultural shocks. So we thought, why not share them with you? Today, we’re continuing our rubric “𝗪𝗵𝗮𝘁 𝘀𝘂𝗿𝗽𝗿𝗶𝘀𝗲𝗱 𝗺𝗲 𝗮𝗳𝘁𝗲𝗿 𝗺𝗼𝘃𝗶𝗻𝗴 𝘁𝗼...”🌍

Our new story comes from Anton, our Senior Software Engineer, who moved to the historic city of Coimbra, Portugal.

Here’s what he says: “In Portugal, the ocean isn't just water, it’s an overwhelming force of nature. Seeing the 50-meter waves in Nazaré makes you realize that the sea is just a puddle compared to the Atlantic. It’s breathtaking and humbling at the same time.”

Swipe through the carousel to see what surprised Anton after moving to Portugal. And stay tuned, more stories from our global team are coming soon!

What surprised you the most after moving abroad? Share in the comments 🙌🏻

Our Meet the Team series, where we introduce you to our colleagues, their hobbies, interests, and life beyond work, continues! 👋

Today, we’d like you to meet Konstantin, Software Engineer at Symfa, who has been with us for about 8 years!

Konstantin started his career as a C # developer and, over the years, has worked on different projects and technologies. What keeps him motivated? New projects, new ideas, and the opportunity to keep learning and put new technologies into practice.

His superpower? Finding effective ways to solve problems 💡

But what does he enjoy outside of work? Which Symfa traditions are his favorite? And what's the best professional advice he has ever received?

Swipe through the carousel to get to know Konstantin better! →

Every two weeks, we host internal webinars where our colleagues share practical AI tips, tools, and work hacks they’ve tested themselves.

One of our recent sessions focused on how to reduce token usage in Claude Code without losing quality. Big thanks to our colleague Anton, Senior Software Engineer, for sharing his insights and practical recommendations! 🙌

Claude Code token usage is split into input and output. Output is what Claude generates, while input includes:

→ your instructions
→ files Claude has read
→ conversation history
→ CLAUDE.md
→ tool results

In long agentic tasks, input can become the main part of the bill. Why?

1. Long conversations create compounding costs

Every new message means Claude processes the conversation history again.

First message → ~500 tokens
30th message → ~15,500 tokens

After 30 messages, cumulative usage can reach around 250K tokens.

2. Bigger context ≠ better results

As sessions grow, context rot can appear:

→ retrieval accuracy: ~92% at 256K → ~78% at 1M
→ reasoning depth can decrease
→ edits without rereading relevant files become more frequent

A 500K-token session can sometimes perform worse than a 200K one, while costing more.

3. Don’t wait for auto-compact

Claude Code auto-compacts at around 95% of the context window.

A practical alternative is to manually compact at 50–70%:

summary → /clear → paste summary → continue

4. Model ≠ Effort

Model = what Claude knows.
Effort = how much work Claude is ready to do.

Subtle bug / unfamiliar domain / architecture → larger Model

Skipped files / tests / verification → higher Effort

Routine task on a large model → smaller Model

And this is only the diagnosis 👀

In Part 2, we’ll move to practice: CLAUDE.md, MCP, Skills, Hooks, and other ways to reduce unnecessary token usage.

During May and June, Symfa specialists took part in an AI Hackathon organized by our client, a large-scale insurance company. We’re proud to have been invited to participate as a technological partner with the client’s internal teams.

Together, we tackled real business challenges, including insurance claim verification, log analysis, and data validation. Every solution was designed to address a practical business need and simplify everyday work.

The value of the hackathon was about quickly turning real business problems into AI prototypes and validating them in the context of real operational challenges faced by the client’s insurance business.

Participants worked in cross-functional teams that included:
✔️ Developers
✔️ Business Analysts
✔️ QA Engineers

The challenge unfolded across three dedicated Hackathon Wednesdays:

🎯 Design Day
Teams explored business challenges, defined requirements, and shaped solution concepts.

💻 Build Day
Ideas turned into prototypes through development, testing, and rapid iteration.

🚀 Test & Demo Day
Teams refined their solutions and presented them to a judging panel.

Symfa was represented by both fully Symfa teams and mixed teams working side-by-side with the client’s specialists.

Our teams had the opportunity to collaborate directly with the client’s teams, combine different perspectives, and shape solutions around real business needs. For our engineers, analysts, and QA specialists, it was also an opportunity to explore new AI tools, step outside their everyday project work.

In our next post, we’ll share what the experience looked like through the eyes of the participants.

teamwork