Hire Brunoo B. - Fine-Tuning Engineer

Designed and developed a multi-agent equity research platform using LangGraph, integrating retrieval, macroeconomic analysis, sentiment analysis, and fact-checking agents to automate complex financial research workflows, reducing analyst effort by 40%. Enhanced RAG pipeline performance by resolving chunking inconsistencies and eliminating noisy data, achieving 30% improvement in retrieval precision and reducing hallucination rates, validated using RAGAS and LLM evaluation frameworks. Architected and implemented MCP servers to standardize communication between LLM agents and external tools, while enabling execution trace observability (branching, retries), improving debugging efficiency and root-cause analysis by 5x. Fine-tuned open-source LLMs using LoRA and RLHF techniques, and deployed scalable inference endpoints with vLLM on GPU infrastructure, delivering cost-efficient, enterprise-grade AI solutions. Optimized LLM inference performance using advanced techniques such as KV caching, speculative decoding, quantization, and GPU memory optimization, resulting in 35% reduction in latency, 40% lower memory utilization, and increased throughput. Built and deployed production-grade ML pipelines on AWS EKS using Kubeflow, KServe, and ArgoCD, ensuring highly available model serving with 99.99% uptime and robust CI/CD integration. Streamlined MLOps workflows by implementing Terraform-based infrastructure and GitOps practices, improving deployment velocity across 10+ ML models while maintaining full experiment tracking and reproducibility using MLflow.

* Developed an AIOps-based predictive analytics platform to detect system failures and anomalies using Random Forest, XGBoost, and LSTM models, improving incident prediction accuracy by 32% and reducing unplanned outages by 28%. * Built scalable data ingestion and processing pipelines using Python, PySpark, and Apache Kafka to handle high-volume log and monitoring data from servers, applications, and cloud environments. * Performed advanced feature engineering on time-series and log data using Pandas and NumPy to identify patterns related to system performance degradation and incident occurrence. * Designed and deployed anomaly detection models to reduce unplanned downtime by predicting incidents before occurrence, improving system reliability and SLA compliance. * Integrated ML models into production using REST APIs (Flask/FastAPI) and deployed them on AWS (EC2, S3, SageMaker) for real-time monitoring and alerting. * Implemented model monitoring, automated retraining pipelines, and CI/CD workflows using Docker, Git, and Jenkins, improving deployment efficiency by 45% and ensuring consistent model performance. * Collaborated with DevOps, Cloud, and Support teams in an Agile environment using Jira, accelerating issue resolution cycles by 30% and delivering data-driven solutions aligned with business KPIs.