Skip to content

Theseus

Read as page

Ruslan Shulga

VP Engineering, AI Platform

New York, NY

VP Engineering at JPMorgan Chase. I lead the AI platform team behind our multi-agent orchestration and hybrid retrieval, plus the custom MCP servers several thousand employees use every day.

Case studiesAsk AIContactEmailLinkedInGitHub

Leads
a team of 7 across AI research, backend and platform
Reach
multi-agent workflows several thousand employees use firm-wide
Gateway
around 8 product teams route through
Impact
cut manual processing on the flagship workflow by about 55%

Revisions

This site is where a team of AI agents shows what it can ship on its own.

RevDescriptionDrawnCheckedTestedApproved
51Show the first screen to six fresh readers; all six answered all three—Lens—Gate
50The count of who approved each sprint is true, and links to the LogForgeLensProbeGate
49Read the crew's rules on the live site, and find one count to set rightQuillLens—Gate
48Link the crew's rules from the home page and the Log, with their recordForgeLensProbeGate
47Show the rules the crew builds under, each quoted from its own fileForgeLensProbeGate

Every name is an AI agent Ruslan directs. Approved: Gate is CI, his rules and Lens's review; Owner is his own call.

22 planks: 22 replaced, 0 original

  • Seated by a sprint
  • Original
All 50 revisions in the Log

Work

  1. JPMorgan Chase & Co.

    Vice President, Senior Software Engineer Lead – AI Platforms & Systems

    Sep 2021 — Present

    Built the multi-agent workflows several thousand employees use firm-wide, on LangGraph with the Claude Agent SDK and OpenAI Agents SDK, and cut manual processing on the flagship workflow by about 55%. Built a model-agnostic AI gateway over AWS Bedrock, Azure OpenAI and the Anthropic API that around 8 product teams route through, a hybrid retrieval pipeline that lifted precision roughly 35%, and the MCP servers that turned sprint-long AI features into a day or two. Lead a team of 7 across AI research, backend and platform.

    Around 8 product teams

    Model-agnostic AI gateway

    AWS Bedrock

    Azure OpenAI

    Anthropic API

    Detail A. Drawn from the summary's own words: the gateway, who routes through it, and what it sits over.

  2. Earth Designs

    Full Stack Web Developer

    Apr 2020 — Sep 2021

    Ran the React and Next.js migration across several production apps, from design through deployment. Built a reusable component library, set up performance monitoring, and mentored two junior engineers.

  3. PPS Capital

    Information Technology Manager

    Jan 2018 — Apr 2020

    Ran IT operations and infrastructure strategy company-wide, and led internal web apps for analytics and reporting. Set up disaster recovery and cybersecurity practices that protected client data.

  4. Earth Designs

    Technology Manager

    Feb 2016 — Jan 2018

    Led API-driven web platforms and microservices, worked closely with the CEO on tech strategy, and set up the team's documentation and onboarding.

Case studies

  • Multi-agent orchestration platform

    Production multi-agent workflows used by several thousand employees firm-wide. LangGraph orchestrates Claude Agent SDK and OpenAI Agents SDK. Each agent has persistent memory and calls internal tools. Irreversible actions pause for a human checkpoint.

    ~55%
    Manual processing cut on the flagship workflow

  • Hybrid retrieval pipeline

    Replaced an embedding-only RAG with a hybrid pipeline: dense plus sparse retrieval, cross-encoder re-ranking, indexed across Pinecone, Weaviate and pgvector depending on the use case. The pieces that mattered weren't the models. They were the eval harness and the index strategy per domain.

    ~35%
    Retrieval precision improvement

  • Internal MCP servers

    Custom Model Context Protocol servers that connect Claude to internal APIs and databases. Features that used to take a sprint now take a day or two.

    1-2 days vs 1 sprint
    Typical feature ship time

  • Model-agnostic AI gateway

    An AI gateway on top of AWS Bedrock and Azure OpenAI, with direct Anthropic API access for newer Claude models. Around 8 product teams route through it for model selection and automatic failover, with cost tracked per team across Claude, GPT-4, Llama and Titan.

    ~8
    Product teams routing through

  • Multimodal document pipelines

    Document and image processing pipelines on Claude Vision and GPT-4V, with Gemini added recently for cost reasons. Handles the first pass of compliance reviews that used to sit with analysts.

    JPMorgan Chase

Projects

  • Theseus

    This portfolio. An autonomous team of AI agents keeps rebuilding it, sprint by sprint, in public.

    GitHub

  • view-json-react

    An accessible, dependency-free React JSON tree viewer with safe inspection, search and virtualization. Open source.

    GitHub

Stack

  • Python
  • TypeScript
  • LangGraph
  • LangChain
  • Claude Agent SDK
  • OpenAI Agents SDK
  • MCP
  • Pinecone
  • Weaviate
  • pgvector
  • AWS Bedrock
  • Azure OpenAI
  • React
  • Next.js
  • Astro
  • Node.js
  • Docker
  • Kubernetes

Education

  1. Harvard University

    Certificate in Computer Science (CS50)

    Jun 2020 — Dec 2020

  2. Kyiv National Economic University

    Bachelor's and Master's, International Economics

    Sep 2010 — Jul 2015

Contact