Serhii Shchoholiev

I'm the Lead Engineer of the Agents team at Hypha — the platform's intelligence foundation. I define how Hypha builds agents: harness, infrastructure, evals, and applied AI research.

serhii.shchoholiev@gmail.com 🤗

Research all →

At Hypha I'm researching agent tool design: how the shape of a tool changes what an agent can do reliably. I share much of this research in the blogs below. Published work:

LongListBench: A Benchmark for Long-List Entity Extraction from Complex Business PDFsKay.ai, July 2026 · Anton Fedoruk, Serhii Shchoholiev, Akhil Mehta, Vishal RohraDOI · Hugging Face · GitHub · PDF
AI-Facilitated Software Project Generation from Natural Language Using Curated Code SnippetsCEUR Workshop Proceedings, ProfIT AI 2025PDF

Writing all (5) →

Building an eval platformOctober 2026
Shared interfaces, reproducible data, cloud runs and scoring. How we built Hypha's eval framework.
How to build efficient Agent ToolsAugust 2026
Tools are user actions, not functions. A 300-run ablation study.
Building Browser Agents for Long-Horizon workVishal Rohra · August 2026 · X
I contributed to the LLM in the Loop (LITL) and coding agents parts of this work.

I post shorter notes on LinkedIn and X.

Elsewhere

Introducing the Agents APIOpenAI, September 2026 · the write-up
Earning customers’ trust is critical in financial services. OpenAI’s Agents API enables us to build more reliable agents, giving customers the confidence to use them in production. By separating the agent harness from the sandbox, we reduced failed agent responses by 86%.
The builder's guide to GPT‑5.6OpenAI, August 2026
Luna keeps 98% of GPT‑5.5's extraction accuracy at one-eighteenth the cost. That gives our agents high-quality document understanding at a price that makes it practical across many more workflows.

Career

Hypha — Lead Engineer, Agents2026 –
Lead Engineer of the Agents team — Hypha's intelligence foundation. Created the agent platform: a harness-agnostic framework with pluggable intelligence, compute, and storage providers. Built a reproducible cloud eval platform with frozen data, PR-specific environments and shared scoring, and the external MCP server.
Related writing & research
- Building an eval platform
- Path to more reliable agents: Migrating to OpenAI Agents API
- How to build efficient Agent Tools
- Quoted in OpenAI's Introducing the Agents API (September 2026).
- Quoted in OpenAI's The builder's guide to GPT‑5.6 (August 2026).
Kay.ai — Founding Engineer2025 – 2026
Built autonomous browser agents for insurance operations, including LLM in the Loop (LITL) recovery and coding agents.
Related writing & research
- LongListBench — benchmark for long-list entity extraction from complex business PDFs; dataset and paper released under doi.org/10.57967/hf/9768.
- Building Browser Agents for Long-Horizon work — contributed to the LITL and coding agents parts.
Claritev — Senior Software Engineer, Lead2022 – 2025
Led a 10-person team automating healthcare workflows; President's Award, 2024.