Open to opportunities

Alejandro palacios

@alejandropalacios

Message

Senior AI engineer specializing in high-performance LLM and multimodal inference optimization.

United States

Message

What I'm looking for

I seek roles focused on scalable, efficient LLM/VLM inference and multimodal systems where I can optimize performance, collaborate with hardware/cloud teams, and deliver measurable cost and latency improvements.

I am a senior AI engineer with about 10 years building production ML systems and high-performance inference at scale, including roles at PayPal, OpenAI, and NVIDIA. I specialize in PyTorch, TensorRT-LLM, quantization, speculative decoding, and GPU acceleration, and have consistently delivered 20–45% performance and cost improvements.

At NVIDIA I built TensorRT-LLM AutoDeploy features and optimized long-context and vision-language workloads; at OpenAI I scaled GPT pipelines and production inference for ChatGPT; at PayPal I developed real-time fraud ML systems. I collaborate across hardware, cloud, and product teams to drive measurable throughput, latency, and safety wins.

Experience

Work history, roles, and key accomplishments

Current

Senior Software Engineer

Current

NVIDIA

Feb 2024 - Present (2 years 1 month)

Built TensorRT-LLM AutoDeploy features and optimizations that reduced manual build time from days to hours and delivered 3–5× higher throughput on long-context decode-heavy workloads; enabled 30–40% lower latency for vision+text tasks and validated stacks on next-gen GPUs yielding 2–4× production throughput gains.

TensorRT PyTorch GPU Acceleration Speculative Decoding Performance Tuning CI CD

Senior Software Engineer

OpenAI

Apr 2019 - Feb 2024 (4 years 10 months)

Scaled end-to-end GPT pipelines and high-throughput inference engines that enabled ChatGPT launch and reduced serving costs ~40% via intelligent batching and runtime optimizations; contributed multimodal prototypes improving image+text reasoning ~22% and built safety/alignment layers reducing harmful outputs ~30%.

PyTorch Hugging Face Distributed Serving Batch Monitoring

Software Engineer

PayPal

Nov 2017 - Apr 2019 (1 year 5 months)

Developed backend services and ML features for real-time payment fraud detection, improving detection accuracy ~25% using behavioral signals, device fingerprinting, and transaction-graph models while contributing to scalable Java/Spring Boot microservices on AWS.