Open to Opportunities

Madhavan Panneerselvam Kumar

I build production-grade agentic systems — RAG pipelines, multi-agent orchestration, and RLHF fine-tuning — from scratch. I differentiate through eval-first development: golden datasets, LLM-as-judge scoring, and CI quality gates that confirm systems work before scaling them.

View Resume GitHub Contact

AI/ML

Specialization

M.S.

Graduate

2026

Class Of

Python✦TypeScript✦React✦Next.js✦LangChain✦LangGraph✦RAG✦RAGAS✦PyTorch✦FastAPI✦PostgreSQL✦Docker✦AWS✦GCP✦Claude API✦OpenAI✦Hugging Face✦Qdrant✦LangSmith✦MCP✦Framer Motion✦Three.js✦Python✦TypeScript✦React✦Next.js✦LangChain✦LangGraph✦RAG✦RAGAS✦PyTorch✦FastAPI✦PostgreSQL✦Docker✦AWS✦GCP✦Claude API✦OpenAI✦Hugging Face✦Qdrant✦LangSmith✦MCP✦Framer Motion✦Three.js✦

01. about

About Me

I specialize in building production-grade AI systems — from RAG pipelines and vector search to full-stack web applications. I care deeply about developer experience, clean architecture, and shipping products that actually work.

M.S. Artificial Intelligence

DePaul University

Class of 2026

02. skills

Tech Stack

🤖

LLM & NLP

Claude APIOpenAI APIPrompt EngineeringLangGraphLangChainMCP Servers

📊

Evals & Observability

RAGASLLM-as-a-JudgeLangSmithCI/CD Quality GatesRegression Testing

🧠

ML & Modeling

PyTorchHugging FaceRLHF (SFT, RM, PPO)scikit-learnXGBoostSHAP

🔍

RAG & Retrieval

RAG Pipeline DesignQdrantpgvectorBM25 + Dense Hybrid

🎨

Frontend

TypeScriptReactNext.jsTailwind CSSFramer Motion

⚙️

Engineering & Infra

PythonFastAPIDockerPostgreSQLGCPAWSGitHub ActionsSQL

03. experience

Work Experience

Machine Learning Intern

Global Techno Solutions

Jun 2023 – Sep 2023India

Built production Python ML pipelines (scikit-learn, TensorFlow) automating ingestion, validation, training, and inference.
Reduced runtime 30% through GPU workload profiling and restructuring.

Data Analytics & Process Automation Intern

Alteryx

Jun 2022 – Aug 2022India

Turned 5+ messy manual workflows into automated, validated ETL pipelines across 3 departments — a 60% effort reduction.
Wrote SOPs so the team could own the pipelines independently.

04. research

Research

ML-Powered Corporate Bond Liquidity Scoring: A Multi-Measure XGBoost Framework on FINRA TRACE Data with SHAP Interpretability

2nd International Online Conference on Risk and Financial Management (IOCRF 2026), JRFM/MDPI

Accepted

Developed an XGBoost-based liquidity scoring model on FINRA TRACE bond transaction data, engineering multi-measure features across bid-ask spread, turnover, and price impact to capture composite market liquidity.
Applied SHAP for model interpretability — identifying which liquidity signals drive predictions most, enabling transparent, auditable scoring.

05. projects

GitHub Projects

NLP-Text-Analysis

Jupyter notebook for NLP text cleaning, tokenization, word frequency statistics, and Zipf’s Law visualization on plain text files.

Jupyter Notebook

Production-RAG-System-with-Evaluation-Framework

A production-grade RAG system (FastAPI + pgvector) combining dense embeddings (BAAI/bge-large) with BM25 sparse retrieval via Reciprocal Rank Fusion. Includes RAGAS evaluation metrics, LangSmith tracing, Redis caching, and full GCP deployment via Terraform + GitHub Actions CI/CD. Endpoints: /ingest, /query, /evaluate.

Python

f1-tracking

Real-time F1 car detection & tracking with YOLOv8 + DeepSORT — Streamlit demo, Dockerized, tested, CI/CD

Python

PKMadhavan

No description provided.

RLHF-Fine-Tuning-Pipeline-SFT-Reward-Modeling-PPO

From-scratch InstructGPT-style RLHF pipeline (SFT, reward modeling, PPO) on GPT-2 in PyTorch, no trl dependency. Diagnoses why the baseline underperforms and tests four fixes (larger model, few-shot prompting, reward-steered decoding, improved RM), reporting real measured results. Packaged with a CLI, pytest suite, and CI.

Python

Dreamer-Autonomous-Driving-RL

Dreamer-V1 world-model RL agent that learns to drive on a highway by training entirely on imagined latent-space rollouts — PyTorch, Docker, CI, live demo

Python

View all repositories

06. resume

Resume

Download PDF Open in New Tab

PDF preview available on desktop.

Download Resume

07. contact

Get In Touch

I'm currently open to new opportunities. Whether you have a question, a project idea, or just want to say hi — my inbox is always open.

GitHub

github.com/PKMadhavan

linkedin.com/in/madhavan-pk

pkmadhavan9802@gmail.com