Computer Science & IoT Undergraduate
PRIYANSHU SHEKHAR
Aspiring AI Engineer & Software Developer
Building high-throughput distributed systems, event-driven architectures, and intelligent AI applications. Currently engineering backend infrastructure at LinkMorph.
Systematic
Problem
Solving.
I am a Computer Science and IoT undergraduate with a strong foundation in software development and an active focus on artificial intelligence. I enjoy building real-world applications, solving complex problems, and continuously expanding my technical knowledge across the full stack.
Core Focus
- AI
- Software Engineering
- Full Stack
Currently Learning
- AI Engineering
- System Design
- Advanced Development
Featured Engineering
A selection of practical systems, machine learning architectures, and performance-critical software engineered from first principles.
Problem
Financial institutions require extremely low-latency, interpretable fraud detection systems that can scale and explain their decisions to analysts.
Solution
Engineered a full-stack, real-time fraud detection pipeline integrating an XGBoost ML model with a FastAPI backend and a Next.js WebSocket dashboard.
Engineering Highlights
- Engineered a real-time fraud detection API using FastAPI, predicting transaction risk scores with sub-second latency using an optimized XGBoost model.
- Implemented a live WebSocket-based Next.js dashboard for analysts to monitor high-risk transactions, visualize risk distributions, and receive instant anomaly alerts.
- Built an automated MLOps pipeline covering synthetic data generation, feature engineering, model training, and SHAP-based explainability for transparent AI decisions.
- Integrated LLM capabilities for automated risk report generation and contextual analysis of suspicious transaction patterns.
- Architected a scalable deployment setup utilizing Docker, Nginx, and GitHub Actions for continuous integration and delivery.
Tech Stack
Problem
Experimenting with multiple ML models and comparing their explainability metrics usually requires extensive custom scripting.
Solution
Built a unified visualization lab that allows rapid experimentation, evaluation, and visualization of 10+ machine learning workflows in a single intuitive interface.
Engineering Highlights
- Built an interactive Streamlit application supporting preprocessing, model training, evaluation, and visualization.
- Integrated 8+ analytical visualizations including confusion matrices, ROC/PR curves, SHAP analysis, and feature importance.
- Designed an intuitive interface for comparing multiple machine learning models, enabling faster experimentation and model selection.
Tech Stack
Problem
Traversing large directory structures and filtering files often consumes excessive memory and time when done inefficiently.
Solution
Engineered a low-memory, high-performance file operations library with streaming I/O and .gitignore-style filtering.
Engineering Highlights
- Engineered a library using C++17 and Python to scan, search, hash, and index thousands of files with low memory overhead.
- Optimized file processing through streaming I/O and ".gitignore"-style filtering, preventing entire files from being loaded into memory.
- Designed a modular, cross-platform architecture with automated testing and robust exception handling.
Tech Stack
Problem
Text-based files require efficient compression to reduce storage without risking data loss.
Solution
Developed a custom binary file format and compression pipeline based on Huffman Coding that achieves 100% data integrity.
Engineering Highlights
- Developed a lossless file compression utility capable of compressing and restoring files with 100% data integrity.
- Implemented a custom binary file format and pipeline that achieves storage savings on text-based files.
- Built a modular C++ application with CMake-based build automation and test cases ensuring portability.
Tech Stack
Built Consistently
A real-time overview of my coding activity, open-source commits, algorithm problem-solving, and engineering consistency.
1000+
Chess Peak Rating
Strategic & Tactical Play
GitHub Contribution Activity
Live contributions from @pinshucodes
Technical Arsenal
A battle-tested stack spanning high-throughput distributed backends, AI systems, secure architectures, and modern web applications.
Languages & Core
Distributed Systems & DBs
Web & Frameworks
AI & Machine Learning
Architecture & Security
DevOps & Cloud Tooling
Experience
Engineering & Startup Roles
Software Engineer Intern
September 2026 - Present- High-Throughput Distributed Systems: Architected a URL redirection engine in Go 1.24 and Redis 7 delivering sub-15ms fast-path 302 redirects (p95: 3.9ms) and handling >7,500 events/sec, using singleflight deduplication to prevent database stampedes under viral concurrency.
- Financial Software & Ledgers: Built an authoritative double-entry ledger with atomic balance debits (SELECT ... FOR UPDATE) and non-negative constraints, eliminating double-spending across 100 concurrent threads and maintaining zero balance drift across 50,000 transactions.
- Event-Driven Architecture: Designed asynchronous click ingestion using Redis Streams consumer groups, decoupling user latency (<0.8ms) from hourly analytics and automated Dead-Letter Queue (DLQ) poison-pill isolation.
- Defensive Application Security: Implemented an SSRF-hardened HTTP client with socket-level net.Dialer.Control hooks blocking DNS rebinding/cloud metadata, enforced custom domain API route isolation, and deployed HMAC-SHA256 anti-tamper verification.
- End-to-End Quality Engineering: Shipped a React 19/TypeScript 5 and Go full-stack platform featuring a 3-tier RBAC admin center, GeoIP routing, and automated test automation spanning 81 Go test files and 131 frontend unit tests with 100% green pass rate.
AI Engineer Intern
Apr 2026 - May 2026- Developed a Retrieval-Augmented Generation (RAG) application using Python, LangChain, FAISS, and Hugging Face to deliver context-aware breast cancer predictions with evidence-backed explanations.
- Built an end-to-end AI pipeline spanning data preprocessing, embedding generation, vector indexing, semantic retrieval, and LLM inference to improve the relevance and interpretability of model outputs.
- Implemented FastAPI backend services integrated with a Streamlit interface, enabling seamless interaction with the RAG pipeline through a user-friendly web application.
- Enhanced response quality through prompt engineering and retrieval optimization, while collaborating with the team using Git/GitHub and code review best practices.
Education
Academic Background
B.Tech in Computer Science and Engineering (IoT)
Class 12 (Intermediate)
Class 10
Have an Idea?
Let's Build Something Interesting.
Always open to discussing new projects, technical challenges, and opportunities in AI & Software Engineering.