Baltic Systems AB guides your organization's AI journey — intelligent agents, deployed safely, in any domain.
Senior consulting in Agentic AI and embedded software, from Stockholm to the world.
Talk to usApply to open positionsAI, out in the open
Everyone's using AI at work. Almost nobody admits it. We think it should be a skill — with guardrails, training, and transparency.
What we doSenior consulting, two deep practices
Agentic AI for any domain — and the embedded engineering depth to run it anywhere, from cloud to device.
Agentic AI, deployed safely
LLM agents and multi-agent systems with guardrails, evaluation, and human oversight. Intelligence you can trust.
RAG & local inference pipelines
Retrieval-augmented generation over your own data, with local inference (Ollama, vLLM) so sensitive data never leaves your walls.
Multi-agent orchestration
LangGraph-based agent teams that debate, critique, and converge: analysts, researchers, traders, risk managers.
Tool recommendation & MCP integration
The right tools for the right agents: function calling, Model Context Protocol servers, custom tooling.
Embedded Linux & Yocto
Board bring-up, kernel, drivers, real-time patches, Yocto distributions. 20+ years of senior engineering.
Containers & Kubernetes
Docker and Kubernetes for AI workloads and edge: reproducible builds, observable pipelines.
IEC 61850 & energy systems
Substation automation protocols and realtime testbeds for intelligent electronic devices.
Data & Power Platform
Power BI dashboards, Power Apps, and Power Automate workflows, including AI-assisted reporting.
Projects & platforms
What we build and run — our own open multi-agent framework, client test tooling, and case studies with honest numbers.
TradingAgents — multi-agent AI framework
Our customized 12-agent LLM framework for collaborative market analysis, built with LangGraph and containerized with Docker. Includes Laya, an optional shadow-assessment layer that independently scores each decision stage.
Damage Report App — AI-assisted delivery
A from-scratch Power Platform demo (Power Apps + Automate + Power BI) proving the method: 88 test cases written before a single screen, 55 passing, every bug documented. 2 hours of human hands, 22 hours of AI effort — honest numbers, no magic.
Desi Bazaar — marketplace, spec-first
A marketplace app for Sweden's Indian and Bangladeshi community. Idea to review-ready spec through 7 adversarial review rounds — one caught a payment-regulation blocker before a single line of production code.
Kubernetes training demo
A hands-on K8s training demo environment — 4 hours of human effort with AI assistance. Proof that a small team can ship working infrastructure tooling fast.
Omicron SCORT test simulator
Software test simulator for Omicron SCORT hardware, built under a larger system design in about two weeks — embedded test tooling for a major client.
IEC 61850 realtime testbed
Rapid proof-of-concept and benchmarking platform for intelligent electronic devices in energy systems.
Siva Borra, Principal Engineer
20+ years building systems where failure isn't an option — from substation automation to AI agents. I run Baltic Systems AB as a specialist consultancy: senior engineering, honest numbers, no layers between you and the person doing the work.
Embedded Linux, Yocto, and real-time systems for energy, medtech, telecom, and automotive. Now applying the same rigor to Agentic AI — RAG pipelines, multi-agent orchestration, and local inference that keeps your data inside your walls.
Real-time testbed for energy IEDs
Challenge: A major energy client needed to benchmark intelligent electronic devices against competitors — fast, with credible numbers.
Contribution: Designed and built a rapid proof-of-concept and benchmarking platform for IEC 61850 devices from scratch.
Outcome: Working testbed delivered in weeks, not months — with head-to-head performance data the client could act on.
12-agent market analysis framework
Challenge: Single-LLM market analysis is shallow and unchallenged — no debate, no risk review.
Contribution: Built a LangGraph multi-agent framework where analysts, researchers, traders, and risk managers debate and converge — containerized with Docker.
Outcome: A working open framework in ~3 hours of hands-on engineering, proving small teams can ship serious AI tooling. Try the live Laya demo · View on GitHub
Trainings & workshops
Practical, hands-on training for teams adopting AI — working code, no hype.
Eliminating hallucinations
RAG patterns, grounding techniques, and evaluation — so your AI tells the truth or says it doesn't know.
Token-budget control
Keep LLM costs predictable: context budgeting, caching strategies, and choosing the right model for the job.
Safe agent deployment
Guardrails, human-in-the-loop patterns, and staged rollouts for multi-agent systems heading to production.
Built like production, from day one
Containerized agents, observable pipelines, private data. This is how every engagement ships.
Illustration — how we build: containerized agents, observable pipelines, private data.
Join us on client assignments
We place senior consultants with our clients across the Stockholm region. These are the profiles we're hiring for right now — apply and let's talk about where you fit.
Embedded Software Engineer (C/C++, Yocto, Linux)
Senior C++ Developer, connectivity & cybersecurity
Agentic AI Engineer (LLMs, multi-agent systems, RAG)
Power Platform Developer (Power Apps, Automate, Power BI)
Senior Embedded Linux Consultant
Let’s talk about your AI journey
One conversation is enough to see where agents can create value — safely.
Talk to us