Skip to content
View lalith0087's full-sized avatar
0 followers · 2 following
  • Hyderabad

Block or report lalith0087

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. Markdown is supported. 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
lalith0087/README.md

Hi, I'm Lalith 👋

Data Science / ML engineer from Hyderabad, India. B.Tech in Computer Science (Data Science specialization), 2025. I like shipping models end to end: tracked experiments, tested code, containers, CI, and honest evaluation, not just notebooks.

Featured projects

Project What it is Highlights
Fraud Detection MLOps Credit-card fraud detection on the real ULB dataset (284,807 transactions, 0.17% fraud) PR-AUC 0.735 · cost-based threshold tuning · MLflow registry with a promotion gate · PSI drift monitoring · FastAPI + Docker + CI · live demo
Telco Churn MLOps End-to-end churn pipeline Optuna tuning · MLflow tracking and registry · SHAP explanations plus a local-LLM narrative endpoint · tuned XGBoost test ROC-AUC 0.841 · live demo
Second Brain GraphRAG Local, private knowledge-graph assistant (Ollama, ChromaDB, NetworkX) Benchmarks GraphRAG against plain RAG, with the null result written up · laptop-safe low-load mode
Local RAG Assistant Chat with any codebase, fully offline AST-aware chunking · multi-repo support · Ollama + ChromaDB + FastAPI
Image Captioning (CNN + LSTM) InceptionV3 encoder + LSTM decoder on Flickr8k Published at ICACTEA-2025 (team lead)
Fast-AgingGAN Face-aging GAN in PyTorch Lightning Trained on CACD and UTKFace

Stack

Languages: Python, SQL · ML: scikit-learn, XGBoost, PyTorch, TensorFlow, SHAP, Optuna · MLOps: MLflow, FastAPI, Docker, GitHub Actions, pytest · GenAI: Ollama, RAG, ChromaDB, GraphRAG

Currently

Building out my portfolio, with a focus on credit risk and fairness, and on measuring RAG systems rather than just building them. Open to entry-level Data Science / ML roles.

Pinned Loading

  1. telco-churn-mlops telco-churn-mlops Public

    End-to-end ML pipeline predicting telecom customer churn — MLflow tracking, model registry, and a FastAPI serving layer

    Python 1

  2. local-rag-assistant local-rag-assistant Public

    Local, API-free RAG assistant that lets you chat with any codebase — AST-aware chunking, Ollama embeddings/LLM, Chroma vector store, multi-repo switching, FastAPI + chat UI.

    Python 2

  3. fraud-detection-mlops fraud-detection-mlops Public

    end-to-end credit-card fraud detection: cost-based thresholds, drift monitoring, MLflow model registry, FastAPI, Docker and CI

    Python 1

  4. IMAGE-CAPTIONING-using-CNN-and-LSTM IMAGE-CAPTIONING-using-CNN-and-LSTM Public

    Image captioning with a CNN (InceptionV3) encoder and LSTM decoder, trained on Flickr8k — includes training/inference scripts and a Flask web demo for uploading an image and generating a caption.

    Python 1

  5. Face-aging- Face-aging- Public

    Fast-AgingGAN: lightweight face-aging GAN in PyTorch Lightning, trained on CACD and UTKFace, with a pretrained generator and an inference script

    Python 1

  6. second-brain-graphrag second-brain-graphrag Public

    Local, private GraphRAG assistant: builds a knowledge graph from your notes with Ollama, ChromaDB and NetworkX, plus a benchmark of GraphRAG vs plain RAG and a low-load mode for laptops.

    Python 1