Mariana Sequeira
LLM Engineer
Porto, Portugal · UTC+0 · 2 years of experience
Mariana focuses on measurable model behavior, reliable retrieval, and production evaluation.
Request a shortlistWhy this profile fits
Evaluation before optimization
Defines datasets, graders, and regression checks so model changes are measured against real failure cases.
Reliable retrieval and tool use
Improves grounding, structured outputs, tool permissions, and recovery behavior for production workflows.
Cost and latency control
Treats task success, response time, and model spend as one production system rather than separate concerns.
Relevant toolkit
Python
LangGraph
Claude API
OpenAI GPT API
PostgreSQLDelivery evidence
Selected work
Support Agent Reliability Harness
Customer-experience SaaSBuilt an offline eval harness that scores a support agent against 320 labeled transcripts on answer accuracy, tone, and policy adherence. Gating prompt and model changes on the suite raised task success from 71% to 92% and made regressions visible before release instead of after a customer complaint.
Hybrid RAG over Compliance Documents
Fintech scale-upDesigned a hybrid retrieval pipeline combining BM25 and dense pgvector search with a cross-encoder reranker over roughly 40k compliance documents. Semantic chunking and metadata filtering lifted retrieval recall@5 from 0.62 to 0.89 and sharply reduced irrelevant citations in generated answers.
Relevant experience
Career timeline
LLM Engineer
Feb 2025–PresentDevlyn (AI-native staffing) · Remote, Porto, Portugal
- Owns agent reliability workstreams for client teams, converting demo prompts into LangGraph agents with validated tool calls, retries, and tracing, then gating every change behind Braintrust eval suites.
Associate AI Engineer
Jul 2024–Jan 2025Aurora Support Cloud (customer-experience SaaS) · Lisbon, Portugal (Hybrid)
- Shipped the first RAG pipeline for a support copilot over roughly 40k help-center articles, using pgvector hybrid search and semantic chunking to reach recall@5 of 0.89.
AI Engineering Intern
Jan 2024–Jun 2024Meridian Data (fintech data platform) · Porto, Portugal
- Prototyped a document-extraction agent that parsed financial PDFs into structured fields using function calling, reaching 88% field-level accuracy on a 500-document test set.
Hiring this role?
See how Mariana fits the LLM Engineer role.
Skill depth
Show experience in context.
Relevant experience by skill
See the complete skill index
Languages
Frameworks
Libraries/APIs
Tools
Paradigms
Platforms
Storage
Other
Education
Formal background
BSc, Informatics Engineering
University of Porto · 2021–2024 · Final thesis on retrieval-augmented question answering, graded 18/20
Credentials
Certifications
Databricks Certified Generative AI Engineer Associate
Databricks · 2024 · Certified
NVIDIA-Certified Associate: Generative AI LLMs
NVIDIA · 2023 · Certified
Communication
Languages
Portuguese
Native
English
Fluent
Spanish
Conversational
More examples
Similar LLM Engineer profiles.
Aravind Balakrishnan
Senior LLM Engineer
Senior LLM engineer, 13 years, specializing in RAG retrieval quality and eval-driven model behavior.
Kacper Wiśniewski
Senior LLM Engineer
Senior LLM engineer who makes model behavior measurable, not anecdotal.
Henrique Vasconcelos
Senior LLM Engineer
Senior LLM engineer, 9 years, specializing in structured outputs and tool calling.