(RAG PIPELINES)
I design retrieval that answers from the client's own corpus — not from a prompt hoping the model remembers.
- Ingestion and parsing
- Chunking strategy
- Embeddings
- Vector stores
- Hybrid search
- Reranking
- Evaluation
Engineering AI-Native Software, Grounded Pipelines, Fitted Integrations. For Data and Business.
Grounded in systems. Fluent in language models.
I’m Laksamana AryaPutra, an AI Native Software Engineer based in Malang. I build AI and full-stack systems — language models, interfaces, and the software that connects them — and fit that work into the business workflows people already use, from the data underneath a feature to the screen in front of them.
(MCP)
Custom Model Context Protocol servers so language models can act on internal systems with a defined contract.
- Custom MCP servers
- Tool design
- Auth and scopes
- Internal APIs
- Error contracts
(LLM ENGINEERING)
Agents and structured outputs with guardrails, measured for cost and latency rather than demoed once.
- Prompt and agent design
- Structured outputs
- Function calling
- Guardrails
- Cost and latency
(DATA ENGINEERING)
The unglamorous half of every AI system: pipelines that stay clean enough for retrieval to mean something.
- ETL and ELT
- Warehousing
- Data quality
- Feature tables
- Lineage for AI
(AI INTEGRATION)
Fitting the model into the workflow that already exists — the CRM, the desk, the night shift — instead of a new tab nobody opens.
- Workflow mapping
- Product fit
- Ops handoff
- Human-in-the-loop
- Change without theatre
(SELECTED WORK)
Retrieving Context, Continuously Connecting Data, Tools and Language Models.
