Machine learning

GPU-accelerated data curation for LLM training.

Managed vector database for production AI applications.

Generates conference presentation slides (Beamer LaTeX PDF and editable PPTX) from a compiled paper with speaker notes and talk script.

Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and distributed checkpointing.

High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate.

Scalable data processing for ML workloads.

Framework for state-of-the-art sentence, text, and image embeddings.

Comprehensive guide for writing systems papers targeting OSDI, SOSP, ASPLOS, NSDI, and EuroSys.

Prompt format rules for generative video and music models.

Write, edit, and adapt PostHog Signals scouts: scheduled agents that scan a project and report findings.

Author continuously-running online evaluations in PostHog AI observability, grounded in real failure modes you've identified.

Building applications with Large Language Models - prompt engineering, RAG patterns, and LLM integration.

Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware.

Fast structured generation and serving for LLMs with RadixAttention prefix caching.

Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency.

Builds generative AI applications on Amazon Bedrock.

Startup-tailored AWS architecture advice for non-agent workloads, adjusted to the company's stage (pre-revenue through Series B+), team size, runway, and…

Activation-aware weight quantization for 4-bit LLM compression with 3x speedup and minimal accuracy loss.

Selects, deploys, and customizes AI models on Amazon SageMaker.

Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP).

Evaluates code generation models across HumanEval, MBPP, MultiPL-E, and 15+ benchmarks with pass@k metrics.

Evaluates NVIDIA Cosmos Policy on LIBERO and RoboCasa simulation environments.

Evaluates LLMs across 60+ academic benchmarks (MMLU, HumanEval, GSM8K, TruthfulQA, HellaSwag).

Provides guidance for automatically evolving and optimizing AI agents across any domain using LLM-driven evolution algorithms.