Engineering
Autonomous Intelligence
& High-Scale Inference
Hi, I'm Tiara Oryza. I architect production-grade LLM infrastructure, multi-agent autonomous workflows, and low-latency inference engines that bridge frontier AI research with real-world enterprise scale.
Interactive shell • Try typing projects or skills
Core Engineering Competencies
End-to-end expertise spanning foundation model adaptation, distributed inference acceleration, and production multi-agent orchestration.
LLM Fine-Tuning & Alignment
Post-training specialization utilizing LoRA, QLoRA, DPO, and RLHF. Experienced with Axolotl, Unsloth, and DeepSpeed for domain adaptation.
High-Throughput Inference
Optimizing inference latency and GPU utilization using vLLM, TensorRT-LLM, continuous batching, PagedAttention, and FP8/AWQ quantization.
Autonomous Agentic Swarms
Designing deterministic multi-agent state-graphs, reflective reasoning loops, external tool use, and automated self-healing workflows.
Hybrid RAG & Vector Search
Enterprise retrieval architectures featuring BM25 + dense cosine fusion, reciprocal rank fusion (RRF), cross-encoder reranking, and HyDE.
MLOps & Cloud Infrastructure
Containerized orchestration with Kubernetes, GPU auto-scaling, model telemetry, distributed cache strategies, and CI/CD pipelines.
AI Evals & Guardrails
Rigorous automated evaluation pipelines for hallucination mitigation, token leakage protection, latency benchmarking, and red-teaming.
Live AI Inference Playground
Test drive the simulated neural engine. Experience real-time token streaming, compute telemetry, and latency metrics.
Featured Systems & Architecture
Curated portfolio of deployed AI systems, open-source architectures, and benchmarked research implementations.
Autonomous Multi-Agent Research Swarm
Hierarchical state-graph orchestrating autonomous agents for deep web research, financial report analysis, and automated factual synthesis.
Low-Latency Distributed LLM Gateway
Custom vLLM routing proxy with automated KV-cache warming, semantic response caching, and dynamic token-rate throttling across multi-GPU nodes.
Hybrid Semantic Search & Knowledge Core
Contextual retrieval platform combining BM25 keyword matching with dense embeddings, reranked via cross-encoders for zero hallucination.
Medical Domain LLM Adaptation
Custom quantized instruction fine-tuning on specialized biomedical datasets with Direct Preference Optimization (DPO) for clinical reasoning.
Automated LLM Evaluation Harness
Continuous evaluation pipeline checking prompt regressions, output drift, safety toxicity scores, and inference cost variations on commit.
Real-Time Event Streaming Hub
Ultra-low latency server-sent events (SSE) broker delivering real-time LLM token streams to thousands of concurrent client interfaces.
Production Technology Matrix
Engineering Experience
A track record of shipping scalable AI solutions, optimizing model serving, and driving system efficiency.
Senior AI Engineer // Systems Architect
[ Leading AI Labs / Enterprise ]Led the architecture of distributed inference pipelines serving 10M+ daily tokens. Engineered custom multi-agent LangGraph flows reducing manual customer underwriting cycles by 78%.
Machine Learning Engineer
[ Autonomous AI Solutions ]Developed end-to-end RAG retrieval pipelines, fine-tuned domain-specific LLMs with LoRA/QLoRA, and reduced GPU cloud expenditure by 40% through quantization benchmarking.
AI Research Fellow / Software Engineer
[ Computer Science & AI Lab ]Conducted research in transformer attention efficiency, deep reinforcement learning, and high-performance numerical algorithms.
Let's Build Intelligent Systems Together
Whether you're scaling an enterprise LLM gateway, architecting an autonomous agent workflow, or looking for an experienced AI engineer for select advisory roles, my inbox is open.
Send an Architecture Inquiry
Fill in your project requirements or proposal below. All transmissions are encrypted.
Message Received
Thank you for reaching out. Tiara Oryza will review your architecture inquiry and respond within 24 hours.