Elliot Slusky — AI / ML Researcher
AI / ML Researcher · San Francisco, CA & Miami, FL · elliot@slusky.com · LinkedIn
Experience
Stanford University — Research Assistant (Jan 2026 – Present)
- Designed and implemented phased inference profiling separating prefill and decode energy attribution via TTFT-based phase boundary detection, revealing decode accounts for 99.7% of total inference energy at ~6.7× lower power draw than prefill.
- Architected a multi-GPU experiment sweep pipeline spanning ~6,000 configurations with fault-tolerant checkpointing, OOM handling, and nested loop ordering to minimize expensive model reloads.
Meridian.AI (Stealth Startup, funded by a16z) — Part-Time AI Developer (Sep 2025 – Mar 2026)
- Worked on evals to benchmark and improve agents.
- Worked on a data ingestion platform for fine-tuning models.
- Instrumented client-side observability to monitor agent behavior in production.
Independent Research (Jun 2025 – Present)
- Conducted independent research on latent reasoning mechanisms in language models and implemented a recurrent looped Transformer architecture inspired by recent work on adaptive computation.
- Experimented with Mixture-of-Experts to detect underutilized experts and prune them, increasing efficiency and reducing memory.
Independent Web3 Development Consulting, San Francisco (Feb 2021 – Mar 2025)
- Managed a team of three developers to create products for clients.
- Generated over $10 million of revenue for clients via applications on Solana.
- Developed applications to interact and trade decentralized financial markets on Ethereum and Solana.
Human Rights Foundation — AI Research Intern (Apr 2025 – Aug 2025)
- Analyzed data on the Venezuelan military to identify key structural, strategic, and personnel insights.
- Built AI-powered tools to vet individuals for security purposes.
- Helped develop systems for instant analysis of authoritarian regimes.
Cookbook.dev, San Francisco — Intern (Mar 2023 – Sep 2023)
- Built minimum viable products (MVPs) and integrated them into production.
- Contributed to Cookbook's smart contract library.
- Created ChefGPT, one of Cookbook's most popular features with over 5,000 monthly users.
Hackathons
- 1st Place — AGI Fight Club (hosted by NFX). Winner of a four-round timed competition building AI applications.
- 3rd Place (out of 50) — Cognition/OpenAI Hackathon. Built "Voice Of The People," instant analysis of political sentiment.
Skills
Machine Learning, Deep Learning, Agentic Frameworks, Rust, JavaScript, Python, Data Analysis, Vector Databases.