This code is used to build & run a Docker container for performing predictions against a Spark ML Pipeline.
A modular framework for secure AI inference, real-time signal processing, and hardware abstraction.
This project demonstrates how parallel reasoning branches, multi-round refinement, and self-consistency voting can dramatically improve reasoning accuracy on a 3B parameter model.
This project is an app that lets the user control the mouse pointer of their computer using their eye gaze. The app uses multiple pre-trained computer vision models from the openvino model zoo in a structured pipeline to detect the user's eye gaze and move the mouse pointer in the right distance and direction.
Using web scraped data from one of India's largest car website, Cardekho, building a prediction service with feature, training and inference pipelines.
Neural network implementations from scratch
WASB-SBDT-FPFilter 是一个可配置、工程化的体育球（网球/足球/羽毛球/排球/篮球）检测与跟踪基线实现，基于 WASB。仓库包含评估代码、示例数据、预训练权重、FP（假阳性）过滤训练与推理工具、交互式 patch 标注工具，并新增了一个一键式端到端推理 Pipeline（WASB 检测 → FP 过滤 → 可视化）。支持通过 Hydra 配置灵活定制，适合研究与工程化部署场景。
A scalable demand forecasting system implementing automated retraining, MLflow-based model management, and cloud deployment using production-grade MLOps practices.
This project is a handcrafted end-to-end Optical Character Recognition (OCR) pipeline built to transcribe my handwritten journal entries into digital text—using PyTorch, Faster R-CNN, AWS Lambda, and iOS Shortcuts. It's a personal and technical showcase of deep learning, MLOps, and full-stack AI deployment. Includes in-depth technical report.
A high-performance, model-agnostic geospatial inference pipeline for gigapixel satellite imagery. Features a memory-efficient chunking engine, flexible batch processing, and a unified CLI/GUI ecosystem for seamless deep learning integration on standard hardware.
Fine-tuned TinyLlama using LoRA on a custom FastAPI Q&A dataset to create a specialized domain expert model. Includes dataset prep, LoRA training script, inference interface, and Colab-ready workflow.
An end to end prediction service for flagging credit card fraud.
🚀 Scale test-time compute with PaCoRe, a framework that enables parallel reasoning through coordinated exploration and multi-round synthesis.
Add a description, image, and links to the inference-pipeline topic page so that developers can more easily learn about it.
To associate your repository with the inference-pipeline topic, visit your repo's landing page and select "manage topics."