Skip to content
View Pradumn13's full-sized avatar
2 followers · 179 following

Block or report Pradumn13

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.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

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

Report abuse
Pradumn13/README.md
Typing SVG

LinkedIn Portfolio Email Buy Me a Coffee

🚀 About Me

I'm a Data Scientist & AI Engineer at Tredence Analytics (IIT Indore, 2023), building production AI systems at the intersection of agentic workflows, distributed backends, and large-scale data. From LangGraph pipelines, to MCP agents with zero-downtime hot-swapping, to distributed URL shorteners hitting 1,497 req/s — I ship systems that work at scale.

What I Do:

  • 🤖 Architect agentic AI systems — LangGraph, RAG pipelines, MCP agents, multi-agent orchestration
  • ☁️ Build and provision AWS infrastructure (Bedrock, OpenSearch, DynamoDB, Lambda) via Terraform
  • ⚡ Engineer high-performance backends with async I/O, sharding, caching, and event-driven pipelines
  • 📊 Instrument systems with Prometheus + Grafana — P50/P95/P99 latency, error rate, throughput
  • 🔧 Deploy containerized stacks on Docker Compose with full observability from day one

Core Strengths: GenAI systems, distributed systems, RAG, agentic workflows, AWS, system design.

💼 Featured Projects

🤖 MCP Hot-Swap Multi-Agent Assistant

Hot-pluggable AI agent platform with runtime tool extensibility via MCP stdio transport

Engineering Highlights:

  • 4 isolated MCP servers exposing 14 tools; static AST scanning catalogs dormant servers without process spawning
  • Zero-downtime hot-swap within a single query lifecycle — 3 verified attach/detach cycles
  • asyncio.gather() parallel dispatch + Redis registry (Hash + Sorted Set + pipeline atomic ops)
  • Param auto-correction heuristic resolving LLM key-name mismatches at zero latency overhead
  • 7-container Docker Compose stack — 6 Prometheus instruments tracking P50/P95/P99, tool call rate, error rate
  • 40+ async pytest tests across 5 fault modes · 100% local on llama3.2 · $0 API cost

Tech: Python, MCP SDK, FastAPI, Ollama, llama3.2, Redis, Docker, Prometheus, Grafana, Streamlit

🔗 Distributed URL Shortener

High-performance distributed URL shortening service with full-stack observability

Engineering Highlights:

  • 1,497 RPS with zero errors across 90,000 requests — sharded across 3 PostgreSQL nodes via consistent hashing (uhashring)
  • Multi-layer caching (Look-Aside + Cache-Null) — 100% cache hit rate, 2.7ms P50, sub-230ms P99
  • Per-shard connection pooling (50 connections × 3 shards) sustaining load under 1,500 RPS
  • 37% throughput improvement (1,094 → 1,497 RPS) migrating to fully async aiohttp + uvloop
  • Redpanda (Kafka-compatible) event-driven analytics pipeline — 20+ metric panels across 16 containerized services

Tech: FastAPI, PostgreSQL, Redis, Redpanda, Docker, Prometheus, Grafana, aiohttp, uvloop

🏢 Experience

Data Scientist @ Tredence Analytics · Bengaluru · June 2023 → Present

Project Impact
Autonomous Deviation Intelligence System — LangGraph, Bedrock, OpenSearch, DynamoDB, Terraform CAPA recommendation: 48 hrs → 23 sec across 13.5K+ deviations · 94% match accuracy · 22% precision lift
Enterprise Text-to-SQL Platform — LangChain, FastAPI, PostgreSQL, BigQuery, Snowflake, Milvus 70% latency reduction · 60% API cost reduction across 500+ daily queries · 4 databases · 5+ LLMs
GraphRAG + Anomaly Detection — Neo4j, spaCy, Isolation Forest, Autoencoders KOL identification across 50K+ physician records · 12% reduction in batch failures

🛠️ Tech Stack

Languages

Python SQL C++ JavaScript

ML/AI & GenAI

LangChain LangGraph Hugging Face PyTorch TensorFlow

Cloud & MLOps

AWS Azure Terraform Docker Kubernetes MLflow

Databases & Streaming

PostgreSQL Redis MongoDB Neo4j Pinecone Redpanda

Monitoring & Tools

Prometheus Grafana FastAPI Apache Spark Airflow

📊 GitHub Stats

GitHub Streak
Contribution Graph

🏆 Competitive Programming

Platform Handle Rating / Rank
⚡ LeetCode Pradumn_89 Rating 1529 · 450+ problems
🏆 CodeForces Pradumn13 Pupil
🍴 CodeChef pradumn_01 Rating 1806 · 3★
📊 Total — 1000+ problems solved

🎯 Current Focus

  • 🔭 Building production AI agents with LangGraph, MCP, and AWS Bedrock
  • 🏗️ Deepening expertise in distributed systems and agentic workflow orchestration
  • ☁️ Expanding AWS + Terraform infrastructure-as-code practices
  • 🤖 Exploring advanced RAG architectures and multi-agent coordination patterns
  • 💬 Ask me about LangGraph, RAG, MCP agents, AWS Bedrock, distributed systems, caching

📫 Let's Connect

LinkedIn Email Portfolio Buy Me a Coffee

💡 "Shipping AI and Distributed Systems that work in production."

Profile Views

Popular repositories Loading

  1. awesome-piracy awesome-piracy Public

    Forked from Igglybuff/awesome-piracy

    A curated list of awesome warez and piracy links

    HTML

  2. mcp-hotswap-agent mcp-hotswap-agent Public

    An AI assistant with hot-pluggable MCP tool servers — one Router Agent, dynamic tool discovery, and LLM-powered routing. Built with FastAPI, Redis, Ollama, Streamlit, and Prometheus.

    Python

  3. distributed-url-shortner distributed-url-shortner Public

    ⚡ High-performance distributed URL shortener achieving 1,497 RPS | FastAPI + PostgreSQL sharding + Redis + Redpanda | Zero errors across 90,000 requests | Production-ready with Prometheus/Grafana m…

    Python

  4. portfolio portfolio Public

    HTML

  5. Pradumn13 Pradumn13 Public