hayder

I work at the intersection of theoretical computer science, artifical intelligence, and systems. I am currently a Principal AI Engineer at Allspice. I am also a visiting researcher at Dr. Kinan Dak Albab's SPACE Lab at Boston University and am part of Dr. Minlan Yu's lab at Harvard University.

[github][scholar]

Background
I have worked at multiple tech companies including Stripe, Bloomberg, and Google. I obtained an M.S. in Computer Science from Harvard with a 4.0 GPA. I received a second M.S. in Cybersecurity from the City College of New York with a 4.0 GPA. My advisor at Harvard was Dr. Minlan Yu and my advisor at CCNY was Dr. Allison Bishop. I was an adjunct professor at Adrian College, where I taught an advanced-undergrad course focused on Systems + ML. I was a teaching fellow at Harvard for CS 145: Networking at Scale. I also designed assignments for Dr. Allison Bishop's graduate-level Adversarial AI course at CCNY.

Advising & Consulting
I take on a small number of advising and consulting engagements. My current focus is teams building agents and agentic workflows, including LLM serving, multi-agent infrastructure, and making these systems reliable in production. To get in touch, email me with "[Consulting]" in the subject line.

Open Source
I love creating open source projects and am passionate about the mission of making the fruits of science and technology accessible to everyone. Below are some of my open source projects.

• Orla is a runtime adaptive execution layer for agentic systems. Orla was awarded a Slingshot grant from the Laude Institute.
• PRP LBF and Cuckoo LBF are the world's first learned Bloom filters with provable security guarantees.
• Pulsar RNG is a cryptographic random number generator that uses pulsar data from NASA and ESA as an entropy source.
• Lilypond is a fun retro game I made, you can play it online here. This was later used in a research project centered around cryptographically verifying the authenticity of speedruns.
• During my Master's at Harvard, I made Cheetah, a P4 library that optimizes queries using programmable switches. It works with the Intel Barefoot Tofino switch, which is sadly no longer in production.

• fast_float library used by GCC, Chromium, Boost, DuckDB, ClickHouse etc. [pull requests]
• Caddy web server. [pull requests]
• Google's P4 PDPI. [pull requests]
• Harvard Library Innovation Lab's Perma.cc. [pull requests]
• A small fix to SGLang.

• Orla: A Library for Serving LLM-Based Multi-Agent Systems. CAIS 2026 (Systems Demonstration Track). [arxiv]
• DRCY: Agentic Hardware Design Reviews. CAIS 2026 (Systems Demonstration Track). [arxiv]
• Adversary Resilient Learned Bloom Filters. AsiaCrypt 2025. [arxiv]
• Cheetah: Accelerating Database Queries with Switch Pruning. SIGMOD 2020. [arxiv] [acm]
• Borg: the Next Generation. EuroSys 2020. [acm] [google]
• SwitchV: Automated SDN Validation. SIGCOMM 2022. [acm]
• cISP: A Speed-of-Light Internet Provider. NSDI 2022. [usenix] [pdf]
• A Cloud-based Content Gathering Network. HotCloud 2017. [usenix] [pdf]
• All Proof of Work but No Proof of Play. CFAIL 2025. [arxiv]

• Context Compaction Theory. [arxiv]
• Whistledown: Combining User-Level Privacy with Conversational Coherence in LLMs. [arxiv]
• Adversarially Robust Bloom Filters: Privacy, Reductions, and Open Problems. [arxiv] [eprint]
• An Introduction to Protein Cryptography. [eprint]
• LSM Trees in Adversarial Environments. [arxiv] [eprint]
• Random Number Generation from Pulsars. [arxiv] [eprint]
• Computational Complexity of Game Boy Games. [arxiv]