Skip to content
View burrows99's full-sized avatar
🏠
Working from home
🏠
Working from home

Block or report burrows99

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
burrows99/README.md
Header

AI and agent engineer in London: multi-agent orchestration, MCP servers, evals

Raunak Burrows

AI and agent engineer in London. I build agentic systems that ship to production: multi-agent orchestration, MCP servers, and the eval pipelines that keep them honest, in Go, Python and TypeScript.

  • 🔭  Currently a tech lead in an AI Products & Platform team
  • 🌍  52 merged open-source PRs across Windmill, Kubernetes Kueue, Hugging Face Transformers, n8n, Google ADK and more
  • 🧭  5+ years in production: clinical AI orchestration, LLM evals for Grok, MCP agents over HIPAA claims, Go microservices for eBay
  • 🌱  MSc AI, University of Surrey (Merit): my dissertation cut NHS doctor wait times by 58% with a Mixture-of-Agents
  • 💬  Ask me about agents, MCP servers, evals or Go microservices
  • 🎹  Off-keyboard: chess, piano, guitar, drums, and a sketchbook
  • 📫  Reach me at raunakburrows@gmail.com

LinkedIn  macquery.in  X  LeetCode  Resume


💼 Experience

Career Gantt chart, newest first: Vice President, AI Products & Platforms, Citi (Nov 2026–now, AI platforms · VP); Senior Engineer, Hesta Health (Apr 2026–Oct 2026, PHI-safe LLM); AI Engineer, Aixolotl (Oct 2025–Sep 2026, MCP · Azure); Data Scientist, Turing (Jun 2024–Aug 2024, +25% relevance); Software Engineer, Sciometrix (Jul 2023–Apr 2025, −65% overhead); Full Stack Developer, QuestionPro (Aug 2022–Jul 2023, NPS 3.8→4.6); Specialist Programmer, Infosys (Sep 2021–Aug 2022, 2× throughput)

Roles and highlights
Role Highlight
Vice President, AI Products & Platforms
Citi · 2026–now
Building AI-powered products, backend services and internal developer platforms for the bank's engineers in the AI Product & Platform team of Citi's Chief Technology Office (Go, Kubernetes, LLM-powered applications)
Senior Engineer
Hesta Health · 2026
Built the clinical AI orchestration service (Python/FastAPI, LangGraph) for a postnatal care platform: parallel DAG execution, per-run model routing, and PHI de-identified with Microsoft Presidio before any payload reaches a Claude model
AI Engineer
Aixolotl · 2025–26
Architected and shipped AIxChat from zero to one: a production AI platform orchestrating multi-step LLM pipelines via MCP, with tenant-boundary enforcement, deployed to Azure
Data Scientist
Turing · 2024
Eval-driven pipelines (SFT, RLHF, LLM-as-a-Judge, golden datasets) for the Grok LLM on X, improving response relevance by 25%
Software Engineer
Sciometrix · 2023–25
An MCP validation agent with 12 tools over HIPAA-regulated claims, cutting validation overhead by ~65% and raising claim accuracy by ~35%; a Go X12 EDI engine saving ~$1,400/month
Full Stack Developer
QuestionPro · 2022–23
FastAPI LLM microservices shipped as Genie AI; post-launch NPS rose from 3.8→4.6/5
Specialist Programmer
Infosys · 2021–22
Go microservices for eBay's retail platform: 2× throughput, latency cut from 4→2 ms with pprof profiling

🌍 Open source

I have 52 merged pull requests across 13 projects in agent, workflow and ML infrastructure. Most of them are bug fixes; Windmill's founder merged 27 of my 30 Windmill PRs, four of them features.

✅ 52 merged · 13 projects
Merged PRs by project: Windmill 30, Kueue 5, Transformers 4, inference-sim 4, llm-d router 1, ADK Go 1, n8n 1, Twenty 1, GPUStack 1, goose 1, marimo 1, ContextForge 1, Stickler 1

🟡 18 in review · 🕰️ 6 earlier
KEDA — in review: Metrics API scaler label matchers (approved) Grafana — in review: unescaped playlist names on delete (approved) CUDA-Q — in review: cudaq.logical.measure traces in a logical program (approved) marimo — in review: data sources search ignores whitespace goose — in review: desktop rejects non-image attachments on external backends MLflow — in review: run-root runs:/ URIs in artifact download and list PostHog — in review: 12 in the morning selects midnight Supabase — in review: case-insensitive database search in Studio Sem. Router — in review: Operator startup probe initialDelaySeconds Qwen Code — in review: mask nested /config secrets, /copy N skips tool-call turns Warp — in review: hash captured constants at full precision ADK Python — in review: failed task delegation answers with the error; same-turn delegations kept apart kafka-ui — in review: bad preview JSON path no longer crashes Messages Karapace — in review: protobuf references for format=serialized Defender — earlier: prevent prototype pollution in object traversal Planner MCP — earlier: built the FastMCP server, OBO auth and Planner service layer Kratix — earlier: contributor-docs fix graphify — earlier: sync Claude references with vocab expansion DebugMCP — earlier: reported the debug-skill link, since fixed Retrieval — earlier: Ollama and Elasticsearch integration +35 more projects, 48 open PRs

Full list with highlights
ProjectMergedHighlights
🆕 Recent: merged in Nov 2026
30 Folder write-permission prefix bug · UTF-8 webhook GET payloads · Clearing a variable's value · features: argument descriptions from JSDoc and Python docstrings, Python client retries on 429 and 502–504, resource usage count per resource type
5 ResourceFlavor that fits the Pod in the Pod importer · Truncated names ending in a dot, backported to release-0.20 and 0.19 at a maintainer's request
4 Multimodal pipelines return the chat as input_text · Docs imports that failed on main
4 Streamed /v1/responses event numbering · Empty top_logprobs in chat logprobs
1 EPP accepts prompt_embeds completions requests
1 Redact and bound code-part labels in agent analytics
1 Discord node keeps the original error
1 API key expiry label in the last 24 hours
1 Reconnected worker keeps its tunnel route
1 Ollama XML tool calls keep <think> tags
1 Variables search stops matching internal cell IDs
1 Translated role labels in the manage-members picker
1 Set a tolerance, not a threshold, in NumericComparator amount examples
🔍 Open and in review
open Metrics API scaler honours !=, =~ and !~ label matchers, after a maintainer assigned the issue · approved
open Stop escaping playlist names in the delete confirmation and button labels · approved
open Let cudaq.logical.measure trace in a logical program · approved
open Data sources search ignores surrounding whitespace
open Desktop rejects non-image attachments on an external backend
open Run-root runs:/<run_id> URIs in artifact download and list
open Picking 12 in the morning selects midnight
open Case-insensitive database list search in Studio
open Apply the Operator startup probe's initialDelaySeconds
2 open Mask secrets nested in a /config parent key value · Skip tool-call-only turns when counting /copy N
open Hash captured constants at full precision
2 open Answer a failed task delegation with the error · Keep same-turn task delegations to one remote agent apart
open A failing preview JSON path no longer crashes the Messages page
open Resolve protobuf references for GET /schemas/ids/{id}?format=serialized
29 more projects 39 open n8n #43942, #43944, #43948 · Mattermost #40050, #40051, #40052 · NocoDB #15046, #15047, #15048 · DeerFlow #8282, #8288 · greplica #108, #109 · HF Chat UI #2770, #2771 · Jupyter Chat #587, #589 · ADK Web #566 · AgentScope #4074 · AI Elements #517 · AWS Labs social messaging toolkit #199 · CopilotKit #8403 · deepagentsjs #951 · Excalidraw #12628 · GPUStack #6796 · Hoppscotch #6796 · Karmada #8065 · Langflow #16430 · LiteLLM #51951 · LiveKit Agents Playground #224 · MCP C# SDK #1922 · Ollama #19707 · OpenHands #19757 · Payload #19603 · Phoenix #17971 · Plane #10287 · Qdrant Web UI #487 · Temporal TS SDK #2695 · WeKnora #5531
🕰️ Earlier
1 Prevent prototype pollution during object traversal
6 Built the FastMCP server, on-behalf-of auth and the Planner service layer
1 Contributor-docs fix
open Sync Claude references with the full vocab-expansion step
issue Reported the README/CHANGELOG link to the debug skill, since fixed
open Ollama and Elasticsearch integration

🏅 Recognition & education

🏆 Recognition

  • 📖 Technical Reviewer of 3D Web Development with Three.js & Next.js (BPB Publications, 2025): chapters 17–18 on performance optimisation and AWS Amplify · post
  • 🥈 2nd place, office hackathon at QuestionPro: a RAG knowledge chatbot with vector embeddings that cut false positives by ~40% · mention

🎓 Education


🧰 My stack

LangChain LangGraph LlamaIndex MCP Hugging Face OpenAI Ollama


📊 By the numbers

Commits Merged PRs PRs opened Issues opened

GitHub Stats Streak

Contribution history graph

Languages by repo Most-committed languages
github-snake

Pinned Loading

  1. nhs-triage-simulation nhs-triage-simulation Public

    Simulation framework for NHS emergency department triage optimization using a Mixture-of-Agents (MoA) architecture — built with SimPy, LangGraph, and FHIR-compliant synthetic data to reduce patient…

    Python

  2. n8n-io/n8n n8n-io/n8n Public

    Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.

    TypeScript 212k 61.6k

  3. huggingface/transformers huggingface/transformers Public

    🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

    Python 169k 35.3k

  4. windmill-labs/windmill windmill-labs/windmill Public

    Code-first platform for workflows, apps and AI agents, on the fastest self-hostable orchestrator (benchmarked against Airflow, Prefect and Temporal). Business teams build with AI, engineers control…

    Rust 18.7k 1.2k

  5. google/adk-go google/adk-go Public

    An open-source, code-first Go toolkit for building, evaluating, and deploying sophisticated AI agents with flexibility and control.

    Go 9k 1k

  6. kubernetes-sigs/kueue kubernetes-sigs/kueue Public

    🚦 Kubernetes-native Job Queueing and Scheduling

    Go 3.2k 927