Fedor Chishchin
AI Engineer & Builder · Senior Full-Stack · 20+ Years Building Software & Companies
fedorstartup@gmail.comfedorthinks.comlinkedin.com/in/teochigithub.com/fedorellogitlab.com/teochiRemote · Uruguay (UTC−3), US time-zone overlap
AI engineer and builder specializing in multi-agent orchestration and specification-driven agent development. Currently a Senior AI Engineer for a US-based biotech client, where I design and implement autonomous AI systems for scientific research — evaluated against multiple public agent benchmarks (incl. FutureHouse's BixBench) and custom internal scenario suites. I operate at the frontier of AI-native development: I don't write code by hand — I direct coding agents (Claude Code on Opus) as the implementation layer while owning architecture, methodology, and quality. 20+ years of engineering give the systems judgment to design what should be built; daily agent practice gives the velocity to ship it.
Core Expertise
- Agent Orchestration
- Multi-agent system architecture · Hierarchical orchestration · Agents-as-tools · Inter-agent protocols · Recursive workflows · ReAct & plan-execute · Typed, recoverable hand-offs with retries and fallbacks
- LLM Engineering
- RAG (chunking, hybrid search, reranking) · Vector search & embeddings (pgvector) · Prompt engineering · Structured output, tool & function calling · Streaming · Context & token-cost optimization · MCP servers (Stdio, SSE, Streamable HTTP) · Model-agnostic — OpenAI, Anthropic, Google, OpenRouter, plus self-hosted / local models via Ollama
- Evaluation & Reliability
- Evaluation frameworks — public agent benchmarks (incl. FutureHouse's BixBench) and private held-out scenario suites · Offline + online evals · Regression suites and guardrails · Tracing, cost control and observability · Quality is measured, not declared
- AI-Native Development
- Specification-driven development — the written spec, not the prompt, is the source of truth · Directing coding agents (Claude Code on Opus) as the implementation layer, end to end · The agents write the code; I own the architecture, the methodology and the quality bar
- Architecture
- System design · API design · Data modeling · Security boundaries · Clean / hexagonal architecture · SOLID · Dependency injection · Event-driven · Performance trade-offs · Decomposition for testability and maintainability
- Stack
- Python (FastAPI, asyncio) · TypeScript · Rust · C/C++ · C# · Node.js · React · Next.js · Vue / Nuxt · PostgreSQL (incl. pgvector) · Redis · Docker · CI/CD · Kubernetes (working knowledge) · GCP — the language is secondary: I don't hand-write code, I own the architecture and the decisions that matter.
Selected Experience
Senior AI Engineer
Mar 2025 — PresentUS-based biotech client · Remote · Confidential
Senior AI Engineer driving the architecture and implementation of autonomous AI systems for scientific research automation.
- Work across the architecture end-to-end: system design, agent orchestration patterns, tool-use design, memory and state management, security boundaries, performance trade-offs — and the implementation that turns architecture into working software.
- Built specification-driven agent development standards that turn a written spec into a working agent rapidly and reproducibly. The methodology compresses the gap between idea and shipped agent and is the reason we move at the pace we do.
- Built the evaluation framework: a measurement methodology spanning multiple public agent benchmarks (including FutureHouse's BixBench) plus custom internal scenario suites — the holdout set the agent never sees during development. The framework, not any single benchmark, is what tells us we're improving.
- Built custom MCP servers (Stdio, SSE, Streamable HTTP) bridging LLMs to proprietary tools and secure execution. Designed RAG pipelines over scientific corpora using PostgreSQL + pgvector. Enforced Clean Architecture, SOLID, DI, and comprehensive testing — directing Claude Code (Opus) as the implementation layer while owning architecture and quality.
Senior Full-Stack & AI Engineer
Feb 2023 — Mar 2025BOTTEC · WowCase · Contento · Independent / Contract
- Built LLM-powered SaaS MVPs from zero as the sole engineer: an AI consultation marketplace (Telegram bot + Flutter app) and Contento (an LLM-driven script generator with prompt orchestration) — design, implementation, integration, release. Separately, on contract, built a Telegram-integrated e-commerce app for BOTTEC with payments, logistics, and warehouse integrations.
- Stack: FastAPI + PostgreSQL + Celery backends, Vue/Nuxt frontends, prompt engineering and LLM integration end-to-end. Also built an AI-augmented multi-strategy trading system in Python + Docker.
Education & Languages
M.Sc. Computer Science · Novosibirsk State University · top Russian CS program
2001 — 2007English professional working proficiency · Russian native
Beyond the CV: 20 years of building businesses — see my ventures →