001 — About

I build things that think.

B.Tech in AI & Data Science at SJCET Palai. I don't just train models — I ship them. Every project in my portfolio went from raw data to a deployed, working product. Reinforcement learning agents. Multi-agent AI platforms. LLM pipelines. All in production.

I'm looking for a team that moves fast and builds real things. Pathanamthitta, Kerala — open to anywhere.

6+
Projects shipped
~80%
RL perf gain
3
AI agents live
2
Industry programs
002 — Skills

The stack.

PyTorch Scikit-learn Stable-Baselines3 HuggingFace PPO / RL CNNs Pandas NumPy Feature Engineering EDA Time-Series SQL A/B Testing Groq API LLaMA 3 Prompt Engineering Multi-Agent Orch. FastAPI Streamlit Vercel SQLAlchemy React 19 Python JavaScript Java Git
003 — Selected work

What I've built.

PROJ — 001

LeafNetwork

Multi-agent AI platform for farmers. Vision, Climate and Satellite agents — real-time crop disease detection, live market signals, AI chat. Shipped independently to Vercel.

React 19 FastAPI PyTorch LLaMA 3
View on GitHub
PROJ — 002

RL Financial Optimiser

PPO agent on financial time-series with custom reward functions and domain-specific state features. Outperformed baseline statistical hedging by ~80%.

PPO Stable-Baselines3 Gymnasium
PROJ — 003

Segmentation Platform

K-Means clustering across 5 behavioural segments with Logistic Regression churn classifier. Live Streamlit inference dashboard for real-time predictions.

Scikit-learn Streamlit Joblib
PROJ — 004

Padik AI

Agentic NLP pipeline transforming raw YouTube transcripts into structured study notes and Q&A pairs via LLaMA 3.1. Zero manual processing post-deployment.

LLaMA 3.1 Groq API NLP
PROJ — 005

E-commerce Analysis

EDA and Linear Regression on 2,000+ customer records. Surfaced $61.27/$38.59/$25.98 annual spend drivers per unit — coefficient-level business insight.

Pandas Scikit-learn Seaborn
PROJ — 006

Neuroevolution Agent

Neural game agent trained via genetic algorithms — gradient-free optimisation, fitness-based selection across generations for real-time decision-making.

PyTorch NEAT Genetic Alg.
004 — Experience

Where I've worked.

British Airways
Feb 2026
Verified
Data Science · Forage
  • Scraped and performed EDA on 1,000+ unstructured customer reviews — surfaced satisfaction drivers across key service dimensions
  • Engineered domain-specific features, trained booking-behaviour classifier, delivered results in executive-facing presentation
Commonwealth Bank
Feb 2025
Verified
Intro to Data Science · Forage
  • Designed end-to-end Python and SQL pipelines for ingestion, cleaning and transformation of large-scale structured financial datasets
  • Applied anonymisation to sensitive customer records; designed relational schemas for dashboard-ready analytical output
005 — Services

What I can do for you.

ML model development

End-to-end pipelines — data prep, training, evaluation, serialised deployment-ready models. Classification, regression, clustering, reinforcement learning.

Data analysis and dashboards

EDA, feature analysis, interactive Streamlit and Plotly dashboards. Stakeholder-ready insight from raw data.

LLM and AI pipelines

Prompt engineering, multi-agent orchestration, unstructured-to-structured pipelines using LLaMA 3 and Groq API.

Full-stack AI apps

FastAPI backends, React frontends, Vercel deployments. Production-ready AI applications built and shipped independently.

006 — Contact

Let's build
something real.

Open to internships, entry-level roles, and select freelance projects. India and remote.

GitHub
shyamhari1074