I'm interested in the space between "AI can do this" and "here is a reliable system that actually does it."

I build intelligent systems at the intersection of AI, machine learning, software engineering, cybersecurity, data, and real-world automation.

My work explores AI/ML systems, retrieval, security, knowledge systems, intelligent software, and robotics — from recommendation engines and geospatial ML to digital forensics and local-first AI.

I care less about collecting technologies and more about understanding the system around them: data, architecture, trade-offs, security boundaries, evaluation, and failure modes.

HARSH-LAB is my central engineering portfolio — a living collection of projects, experiments, technical explorations, case studies, and proof of implementation.

The portfolio shows the work. The repositories show the implementation.

🌐 Explore HARSH-LAB → 💻 View the source →

HARSH-LAB

BUILD → MEASURE → INVESTIGATE → ITERATE → SHIP

Projects vary in maturity. I distinguish between experiments, prototypes, working implementations, and production-oriented systems rather than presenting every repository as production software.

Geospatial machine learning for infrastructure inspection.

Explores risk prediction, uncertainty, fairness-aware prioritization, and active learning.

Python XGBoost Geospatial ML

→ Explore repository

Local-first AI for self-discovery and career exploration.

Explores AI-assisted guidance and privacy-preserving software through a local-first desktop architecture.

React TypeScript Rust Tauri

→ Explore repository

A multi-stage recommendation system.

Candidate Generation
        ↓
Vector Retrieval
        ↓
Learning-to-Rank
        ↓
Re-ranking
        ↓
Diversity Optimization

Explores Two-Tower retrieval, FAISS vector search, XGBoost learning-to-rank, and MMR-based diversity optimisation.

Python PyTorch FAISS XGBoost Docker

→ Explore repository

I think of AI as one layer inside a larger system.

Models
  +
Data
  +
Retrieval
  +
Memory
  +
Knowledge
  +
Tools
  +
Security
  +
Interfaces
  +
Evaluation
  =
Useful Systems

The difficult engineering problems often exist in the connections between models, data, tools, users, security boundaries, evaluation, and real-world constraints.

I prefer building from first principles rather than stopping at a model, tutorial, or attractive demo.

Problem
   ↓
Requirements
   ↓
Research
   ↓
Architecture
   ↓
Implementation
   ↓
Testing
   ↓
Evaluation
   ↓
Failure Analysis
   ↓
Iteration
   ↓
Hardening
   ↓
Shipping

When building systems, I ask:

• What problem is actually being solved?
• What assumptions does the system depend on?
• What happens when the model is wrong?
• Where are the trust boundaries?
• Can the output be explained or traced?
• How should performance be measured?
• What happens outside the happy path?

A working demo is useful. A measurable, reproducible, and understandable system is better.

AI / ML LLMs RAG AI Agents PyTorch XGBoost scikit-learn NLP

Data / Retrieval Python Pandas SQL FAISS Elasticsearch Vector Search

Systems FastAPI Docker Kafka REST APIs OAuth2

Knowledge / Security Neo4j Knowledge Graphs Cybersecurity Digital Forensics Threat Intelligence

Frontend React Next.js TypeScript JavaScript

Robotics / IoT Arduino ESP32 Raspberry Pi Sensors Automation

I complement practical engineering with structured learning across AI, data, software, cybersecurity, and product development through Google and IBM learning programs.

Key areas include:

Artificial Intelligence Data Analytics Cybersecurity Python Automation UX Project Management

🎓 View verified credentials on Credly →

Certifications demonstrate structured learning. Projects demonstrate application.

Alongside software engineering, I enjoy practical technology education involving:

Arduino ESP32 Raspberry Pi Sensors Automation Python AI Robotics

I particularly enjoy taking difficult technical concepts and turning them into things people can understand and build.

Production AI Systems · Agentic Architectures · Retrieval & Knowledge Systems · AI Security · ML Evaluation

I'm interested in:

Applied AI · Machine Learning Systems · AI Engineering · AI Security · Knowledge Graphs · Retrieval · Recommendation Systems · Robotics · Intelligent Software

🧪 HARSH-LAB 💻 GitHub 💼 LinkedIn 🎓 Credly

Don't just say you know it. Build it.

Don't just build it. Measure it.

Don't just measure it. Understand why it works — and why it fails.

Then iterate.

Build systems. Understand the fundamentals. Make the complexity useful.