11/11/2026 updated
AS
100 % available
Tech Lead - Backend & Data Engineering
Sundbyberg, Sweden
Sweden
Master's Degree in Applied Mathematics, KTH Royal Institute of Technology, StockholmAbout me
Tech lead and platform engineer in Stockholm. I like building systems that take tedious work off people’s hands, from cloud platforms to AI and data pipelines. I plan carefully, explain hard things simply and enjoy being the link between developers and management.
Google BigQueryCloud EngineeringNvidia CUDAData ArchitectureNode.jsAzure Machine LearningDatabase EnginesGoogle CloudCloud Platform SystemFeature EngineeringEvent-driven ArchitectureGoogle FirebaseBackendFastAPIEvent Driven Architecture
Backend & Architecture
Expertise in Node.js, Express, Fastify, Flask/FastAPI, REST API design (OpenAPI/Swagger), event-driven architecture, message queues (RabbitMQ), microservices, multi-tenant SaaS, JWT/RBAC, OAuth 2.1, and WebSockets. Architecture ownership end-to-end for a multi-tenant financial-operations SaaS on Google Cloud.
Cloud, Platform & CI/CD
Deep hands-on experience with Google Cloud Platform (GKE, App Engine, BigQuery, Cloud Storage, Cloud Logging), Kubernetes (Helm, KEDA autoscaling, StatefulSets, CronJobs), Docker, Terraform, CircleCI, nginx, API Gateway, Prometheus, Linux/systemd, VPC private peering, and AWS (EC2, S3).
Data & ML Engineering
Queue-driven ETL, scheduled and batch pipelines, financial reporting engines, web crawling and data acquisition, data cleaning and schema normalisation, Pandas/NumPy, LLM fine-tuning (PEFT/LoRA, 4-bit quantisation, HuggingFace, PyTorch/CUDA), self-hosted model serving, embedding/semantic search, Model Context Protocol (MCP), OpenAI API, scikit-learn, XGBoost, forecasting and regression, and feature engineering.
Programming Languages
Proficiency in TypeScript, JavaScript (ES6+), Python, SQL, Bash/Shell, Java, R, and C/C++.
Databases & Data Modelling
Extensive experience with MongoDB and MongoDB Atlas (Mongoose, aggregation pipelines, indexing and query optimisation, change streams), PostgreSQL, MySQL, SQL Server, Firebase Realtime Database, and versioned database migrations.
Engineering Practices
Git version control (Bitbucket, Conventional Commits, commit hooks, branch-based CI/CD, worktrees), test automation (Vitest, Jest, Selenium, Playwright), CI test parallelisation and sharding, migration and contract testing, static analysis (custom Semgrep rules, ESLint), secure coding, code review, Agile/Scrum and JIRA workflow design, spec-driven and agent-assisted delivery.
Leadership & Team Management
Four years leading a 5-6 developer team across three countries (Sweden, Bosnia, Vietnam), owning end-to-end delivery from requirements and architecture through implementation, deployment, and production operations.
LLM Fine-Tuning & Production AI
Fine-tuned and quantized open-weight LLMs (Mistral-7B) using PEFT/LoRA with 4-bit quantization, HuggingFace transformers, TRL SFT, PyTorch/CUDA, and embedding/retrieval layers for hierarchical multi-label classification problems. Served models on self-hosted GPU infrastructure behind a hardened public API.
FinOps & Cost Optimisation
Drove FinOps across GCP estates by cutting Cloud Logging spend, right-sizing queue, backend, and frontend workloads, and scaling idle Kubernetes clusters with Terraform and KEDA scale-to-zero design.
Expertise in Node.js, Express, Fastify, Flask/FastAPI, REST API design (OpenAPI/Swagger), event-driven architecture, message queues (RabbitMQ), microservices, multi-tenant SaaS, JWT/RBAC, OAuth 2.1, and WebSockets. Architecture ownership end-to-end for a multi-tenant financial-operations SaaS on Google Cloud.
Cloud, Platform & CI/CD
Deep hands-on experience with Google Cloud Platform (GKE, App Engine, BigQuery, Cloud Storage, Cloud Logging), Kubernetes (Helm, KEDA autoscaling, StatefulSets, CronJobs), Docker, Terraform, CircleCI, nginx, API Gateway, Prometheus, Linux/systemd, VPC private peering, and AWS (EC2, S3).
Data & ML Engineering
Queue-driven ETL, scheduled and batch pipelines, financial reporting engines, web crawling and data acquisition, data cleaning and schema normalisation, Pandas/NumPy, LLM fine-tuning (PEFT/LoRA, 4-bit quantisation, HuggingFace, PyTorch/CUDA), self-hosted model serving, embedding/semantic search, Model Context Protocol (MCP), OpenAI API, scikit-learn, XGBoost, forecasting and regression, and feature engineering.
Programming Languages
Proficiency in TypeScript, JavaScript (ES6+), Python, SQL, Bash/Shell, Java, R, and C/C++.
Databases & Data Modelling
Extensive experience with MongoDB and MongoDB Atlas (Mongoose, aggregation pipelines, indexing and query optimisation, change streams), PostgreSQL, MySQL, SQL Server, Firebase Realtime Database, and versioned database migrations.
Engineering Practices
Git version control (Bitbucket, Conventional Commits, commit hooks, branch-based CI/CD, worktrees), test automation (Vitest, Jest, Selenium, Playwright), CI test parallelisation and sharding, migration and contract testing, static analysis (custom Semgrep rules, ESLint), secure coding, code review, Agile/Scrum and JIRA workflow design, spec-driven and agent-assisted delivery.
Leadership & Team Management
Four years leading a 5-6 developer team across three countries (Sweden, Bosnia, Vietnam), owning end-to-end delivery from requirements and architecture through implementation, deployment, and production operations.
LLM Fine-Tuning & Production AI
Fine-tuned and quantized open-weight LLMs (Mistral-7B) using PEFT/LoRA with 4-bit quantization, HuggingFace transformers, TRL SFT, PyTorch/CUDA, and embedding/retrieval layers for hierarchical multi-label classification problems. Served models on self-hosted GPU infrastructure behind a hardened public API.
FinOps & Cost Optimisation
Drove FinOps across GCP estates by cutting Cloud Logging spend, right-sizing queue, backend, and frontend workloads, and scaling idle Kubernetes clusters with Terraform and KEDA scale-to-zero design.
Languages
EnglishFluentBosnianFluent
Project history
Productionised from KTH MSc thesis at E-avrop. Fine-tuned and quantized an open-weight LLM (Mistral-7B) using PEFT/LoRA with 4-bit quantization for hierarchical multi-label EU procurement classification. Built the training corpus by crawling 645,000 documents. Served the model on self-hosted GPU infrastructure behind a hardened public API with Google Cloud API Gateway, nginx reverse proxy, and automated infrastructure tests.
Owned architecture and platform decisions for a multi-tenant financial-operations SaaS running around 12 customer companies and roughly 10,000 invoices/month. Led a 5-6 developer team across three countries, co-designed the JIRA/agile process, re-architected tenant imports into isolated per-tenant Kubernetes queue workers, owned the MongoDB data layer end-to-end, built the release-safety migration framework, architected the accounting-integration engine with Fortnox, maintained the analytics layer on MongoDB aggregation pipelines, owned the batch-analytics and AI reporting stack, and opened the platform to AI agents via a Model Context Protocol server.
Worked on two concurrent projects alongside the AccuraSee tech-lead role: Byggkollen (diagnosed and fixed head-of-line blocking under peak load, kept a long-lived event-driven Firebase platform healthy, owned CI and AWS/GCP upkeep plus BigQuery log analytics) and Prism (built a governed self-service data platform with an identity-aware read-only engine over SQL view layers, AI chat clients, Python/FastAPI, Model Context Protocol over Streamable HTTP, fail-closed OAuth 2.1/Microsoft Entra, SELECT-only SQL Server guards, hash-chained append-only audit, Alembic, React 18 portal, Bicep/Docker).