Technical Projects
A showcase of engineering solutions, AI models, and data pipelines focusing on measurable impact.
Featured Case Study: ChainReaction
Architecting Resilient Supply Chains with Multi-Agent AI
The Challenge: Global supply chains face increasing volatility from geopolitical shifts. Traditional risk monitoring is reactive, often missing the "first domino" in a sequence of events.
The Solution: Built using LangGraph and Neo4j, ChainReaction is a multi-agent system that autonomously bridges the gap between global news and specific supply chain nodes using Named Entity Recognition (NER).
The Impact: Successfully demonstrated a 40% reduction in risk response time during trials, providing stakeholders with actionable intelligence before disruptions hit.
Read Full Case Study
EduSched
EduSched
Problem: Educational institutions struggle with multi-constraint scheduling (rooms, instructors, student requirements).
Solution: A hybrid heuristic + Google OR-Tools solver that generates optimal schedules in seconds.
Impact: Reduced administrative scheduling time by ~80% in pilot tests.
View RepositoryChainReaction
ChainReaction
Problem: Global supply chains are vulnerable to rapid, unforeseen geopolitical shifts.
Solution: Multi-agent system (LangGraph) that maps news to supply chain graphs via Neo4j.
Impact: Real-time risk detection allows for 40% faster contingency planning.
View Repository
Prompt Trainer
Prompt Trainer
Problem: AI-generated feedback often lacks consistency and doesn't align with human rubrics.
Solution: Human-in-the-loop system that uses DSPy to programmatically optimize prompts based on user feedback.
Impact: Increased AI grading alignment with human instructors from 65% to 92% over 50 iterations.
View Repository
Vibe Coder
Vibe Coder
Problem: Switching between IDE and AI chat interfaces breaks developer flow state.
Solution: A provider-independent CLI assistant with 40+ specialized commands for refactoring and testing.
Impact: Benchmarked a 30% reduction in typical "chat-to-code" cognitive context switching overhead.
View Repository
Tech-Pulse
Tech-Pulse
Problem: Information overload makes it difficult to track market sentiment and emerging tech trends.
Solution: Real-time Streamlit dashboard using BERTopic for grouping and VADER for sentiment scoring.
Impact: Provides actionable daily intelligence summaries in < 5 minutes of review.
View Repository
ImageMD
ImageMD
Problem: Extracting structured data from complex PDFs with tables and formulas is notoriously difficult.
Solution: Pipeline using vision transformers to convert document screenshots directly into clean Markdown.
Impact: High-fidelity extraction of academic papers for ingestion into RAG pipelines.
View Repository