Hi, I'm

Manas Choudhary

AI Systems Engineer

AI Engineer @ Atlan | Research Intern @ IIT Roorkee | Google DeepMind Open Source Contributor | Building production AI agent infrastructure, LLM bias mitigation systems, and open-source governance tools.

LinkedIn Journey
OpenPeeps character illustration
// NEXT: PROJECTS
Agent OS
RustFastAPIvLLMDocker ComposeTimescaleDBpgvectorRedis

Agent OS

Self-Hosted "AWS for AI Agents"

Point any LangChain / LlamaIndex / OpenAI-SDK agent at one governed port and toggle on only the capabilities you want — governance, memory, GPU optimization, sandbox, skill-mining. Eight integrated layers: Govrix Scout (Rust governance proxy), MemoryMesh (RAG + knowledge), InferCraft (KV-cache engine), Crucible (execution sandbox), SkillForge (skill mining), polyglot Storage, and bundled Observability.

Source CodeLive Demo
Govrix Scout
RustReactTimescaleDBDockerAI Governance

Govrix Scout

AI Agent Governance Platform

Open-sourced a Rust reverse proxy for AI agents with under 5ms p99 latency. Features PII detection, SHA-256 Merkle-chain audit trail, per-agent cost attribution, kill switches, and budget enforcement. Supports OpenAI, Anthropic, MCP, and A2A protocols. 70+ GitHub stars. SOC 2, HIPAA, and EU AI Act compliance-ready.

Source CodeLive Demo
Google DeepMind – Concordia
PythonMulti-Agent AIAI SafetyOpen Source

Google DeepMind – Concordia

Open Source Contributor · 15 PRs Authored

Authored 15 pull requests to DeepMind's generative agent simulation framework (1.5k+ GitHub stars) for AI safety research, spanning a structured-logging migration across 26 files, crash-level bug fixes, and validation and robustness improvements to agent execution and reasoning modules.

ChunkScope
Next.jsFastAPIpgvectorLiteLLMRAG

ChunkScope

Visual RAG Pipeline Builder

Drag-and-drop RAG pipeline builder with AI-powered suggestions, live RAGAS evaluation, and side-by-side comparison. Supports 8 chunking strategies, hybrid retrieval (BM25, HyDE, MMR), cross-encoder reranking, and Python code export. Analyzes document structure to recommend optimal pipeline configuration.

Source CodeLive Demo
Aura
Claude Agent SDKPythonChromaDBPlaywrightTypeScript

Aura

Figma-to-React Generation Agent · Samsung PRISM

An AI agent that converts Figma designs into production React + TypeScript. A single Claude Agent SDK supervisor writes the code, then self-corrects through a build-verification loop and a Playwright + Claude Vision visual loop until the render matches the design. Adds ChromaDB component reuse and optional GitHub/Vercel deploy.

Source CodeLive Demo
Strategic Code Companion
TypeScriptAI AgentsASTBabelD3.js

Strategic Code Companion

AI-Powered Code Analysis Platform

VS Code extension with 4 specialized AI agents (Refactor, Architect, Librarian, Tutor) using multi-dimensional scoring for intelligent chunk selection. Analyzes business flows, architectural patterns, and quality hotspots with D3.js visualization and PDF reporting. The project that secured the Atlan internship.

AI Career Platform
PythonReactFastAPIGeminiLangChain

AI Career Platform

Intelligent Career Matching

Full-stack career development platform featuring 6 specialized agents (CareerAnalyst, MentorBot, SkillsAssessor, LearningOrchestrator, OpportunityScout, ProgressMonitor). Implements Big Five & RIASEC assessments, career fit scoring, context-aware mentor chat with sentiment analysis. React 19 + TypeScript frontend with FastAPI backend.

Technologies I Master
Tech iconTech iconTech iconTech iconTech iconTech iconTech iconTech iconTech icon

TECH UNIVERSE

PythonPython
JavaScriptJavaScript
TypeScriptTypeScript
C++C++
HTML5HTML5
CSS3CSS3
ReactReact
Next.jsNext.js
TailwindTailwind
Three.jsThree.js
FramerFramer
Node.jsNode.js
ExpressExpress
FastAPIFastAPI
FlaskFlask
GraphQLGraphQL
TensorFlowTensorFlow
KerasKeras
LangChainLangChain
CrewAICrewAI
RAGRAG
LLMLLM
LoRALoRA
MCPMCP
MongoDBMongoDB
FirebaseFirebase
Neo4jNeo4j
RedisRedis
AWSAWS
GCPGCP
DockerDocker
GitGit
GitHubGitHub
VercelVercel
React NativeReact Native
ExpoExpo

EXPERIENCE

6 internships by 4th semester. AI engineer who ships production systems.

Atlan logo

AI Engineer

Atlan

Jan 2026 - Present | Singapore HQ · Remote

Building enterprise-grade AI agent infrastructure using MCP (Model Context Protocol) with LangChain and LangGraph. Designed stateful context management with Redis-backed caching across multiple LLM providers, deployed on Docker and Kubernetes. Hardened async tool-call pipelines against race conditions to improve reliability and recovery, with end-to-end observability via Langfuse and LangSmith.

Skills: MCP, LangChain, LangGraph, Docker, Kubernetes, Redis

OPEN SOURCE

Loading GitHub stats...