Chintan Patel

Applied AI/ML Engineer · Machine Learning · NLP · GenAI/RAG

I build end-to-end AI systems that connect reliable data, evaluated models, retrieval pipelines, typed APIs, usable interfaces, human review, and operational evidence.

📍 Calgary, Alberta, Canada  ·  🎯 Open to new-graduate and junior opportunities across Canada

Computer Science and Engineering graduate with more than three years of software-development and client-delivery experience, plus a completed Post-Diploma Certificate in Integrated Artificial Intelligence from SAIT.

I focus on applied AI systems where outputs must be explainable, evidence-backed, reviewable, and honest about uncertainty and limitations. My work spans machine learning, NLP, RAG, APIs, product interfaces, testing, containerization, and deployment.

Target roles: Applied AI/ML Engineer · Machine Learning Engineer · GenAI/RAG Engineer · NLP Engineer · AI-Focused Software Developer · Junior MLOps Engineer

A clinical decision-support workflow connecting clinician notes and structured intake to ESI 3/4/5 prediction, explicit safety escalation, clinician decisions, audit evidence, and PDF reporting.

• Final LightGBM V2 registry with 273 ordered runtime features
• Clinical NLP with editable fields, evidence snippets, safety cues, and missing-data warnings
• Explicit ESI 1/2 safety escalation outside the classifier
• Clinician accept, override, and needs-review workflows with audit history
• React/TypeScript frontend, FastAPI backend, SQLAlchemy persistence, and automated tests

Python LightGBM FastAPI React TypeScript SQLAlchemy ReportLab

Scope: Educational portfolio decision support. It is not diagnostic, clinically validated, or a replacement for clinician judgment.

A local release-style RAG system designed around durable document identity, evidence sufficiency, page citations, controlled rejection, provider failure, and operational readiness.

• SHA-256 duplicate prevention and PostgreSQL document-lifecycle metadata
• SentenceTransformer embeddings with ChromaDB retrieval
• Calibrated answerability, evidence gating, page citations, and unsupported-question rejection
• Provider-resilient citation-only fallback
• FastAPI, Next.js, SQLAlchemy, Alembic, request IDs, structured logs, and readiness checks
• 230 backend tests, 128 frontend tests, and a controlled 16-case evaluation workflow

The verified run intentionally disabled generation. It validates retrieval, answerability, unsupported-question handling, expected-page retrieval, and citation-only fallback—not generated-answer quality or production accuracy.

Python FastAPI Next.js PostgreSQL SentenceTransformers ChromaDB Docker Compose

Scope: Verified local release. No claim of production authentication/RBAC, multitenancy, managed cloud operations, commercial adoption, or generated-answer accuracy.

An NLP platform that separates parsing, role classification, keyword evidence, semantic job matching, skill gaps, writing quality, and reviewer workflow instead of presenting one unexplained hiring score.

• PDF, DOCX, and TXT parsing with normalization and structured review
• TF-IDF role classification plus keyword and semantic job-description matching
• Skill-gap analysis, writing guidance, batch comparison, and reviewer notes
• Privacy-safe display mode and human-reviewed recommendations
• Streamlit application with optional FastAPI, SQLite, SQLAlchemy, Docker, testing, and CI foundations

Python NLP scikit-learn Streamlit FastAPI SQLAlchemy Docker Azure App Service

Scope: Human-reviewed decision support. It does not automate hiring decisions or present model output as employment truth.

A technical assignment evolved into a portfolio project that separates natural-language interpretation from verified business rules.

• Focused Google ADK agent with tool-based product search
• Deterministic Python filtering for category, price, name, and availability constraints
• FastAPI service, React/Vite interface, Docker packaging, tests, and structured error handling
• Google Cloud Run backend and Netlify frontend deployment architecture

Google ADK Python FastAPI React Vite Docker Google Cloud Run

Scope: Portfolio-scale workflow with a small validated catalogue; not a commercial recommendation engine.

• What It Would Cost to Run PolicyGPT for 10,000 Queries
• Building TriageAI's ESI 3/4/5 Model: Why LightGBM Won
• Designing a Review-First Clinical NLP Safety Layer
• How I Evaluate a RAG System Beyond Demo Questions
• From Notebook to FastAPI: Building a Reproducible Model Registry
• Designing Drift Monitoring for Production ML Systems

• Building RegImpact AI, a regulatory-change intelligence and human-review workflow
• Deepening agentic workflows, document versioning, change detection, background processing, RBAC, observability, and cloud deployment
• Preparing for new-graduate and junior Applied AI/ML opportunities across Canada

I am open to new-graduate and junior opportunities across Canada in Applied AI/ML Engineering, Machine Learning Engineering, GenAI/RAG, NLP, AI-focused Software Development, and Junior MLOps.