Leander Francis Fernandes

AI Application Engineer & Full-Stack Developer
Building LLM, NLP & Computer Vision systems with real-world impact.

Resume GitHub LinkedIn Email
Leander Fernandes avatar

About

AI Application Engineer & Full-Stack Developer building production-grade LLM, NLP, and Computer Vision systems.

I design scalable AI applications using Python, FastAPI, and React, focusing on clean APIs and reliable system architecture.

My work includes prompt engineering, vector search (RAG), and real-time computer vision pipelines optimized for performance.

I care deeply about clarity, maintainability, and UX — building systems that are easy to use and easy to evolve.

Skills

Frontend

React
HTML5
CSS3
JavaScript
Tailwind

Backend & APIs

Python
FastAPI
Flask
Node.js
Express

AI / ML

LLMs
NLP
Computer Vision
RAG
Scikit-learn

Data & DevOps

MongoDB
MySQL
Docker
GitHub Actions
Vercel

Projects

Selected projects demonstrating applied AI systems, full-stack development, and production-ready engineering.

LLM Playground Production LLM Application · Dec 2025

Industry-ready LLM playground demonstrating real-world LLM integration with a FastAPI backend, Google Gemini, and a clean React + Tailwind UI, deployed with production-grade architecture.

Python 3.13 FastAPI Google Gemini React Vite Tailwind CSS Pydantic v2 Python Logging Render Vercel
Live Demo GitHub
  • Designed an API-first FastAPI backend with versioned endpoints (/api/v1/ask) for LLM inference and structured responses.
  • Integrated Google Gemini securely on the server side with environment-based configuration, token usage tracking, and centralized error handling.
  • Built a minimal, industry-style React UI with animated response reveals, Markdown rendering, copy-to-clipboard support, and prompt history.
  • Implemented advanced UX features including a token slider (up to 100k), dynamic prompt examples per session, and fully responsive layouts.
  • Deployed backend on Render and frontend on Vercel with production-ready CORS configuration, structured logging, and safe secrets management.

AI-Powered Legal Decision Support System AI / NLP System · May 2025

AI-driven legal intelligence system that predicts relevant IPC sections, estimates punishment ranges, and retrieves semantically similar past cases to support data-driven judicial decisions.

Python Flask React.js SentenceTransformer NLTK Scikit-learn ChromaDB Gemini API Pandas NumPy PyPDF2 GitHub
GitHub
  • Built an end-to-end AI system to analyze crime descriptions and automatically classify applicable Indian Penal Code (IPC) sections using NLP and semantic embeddings.
  • Implemented punishment estimation logic aligned with historical judicial precedents and retrieved similar past cases via vector search using ChromaDB.
  • Integrated Google Gemini API to generate concise, context-aware summaries of retrieved legal cases for faster interpretation.
  • Designed the system to handle first-person, third-person, and mixed legal narratives, achieving up to 77.63% IPC classification accuracy.
  • Developed a React-based frontend to present predictions, case matches, and AI-generated summaries in an interactive and readable format.

CineVault CineVault = Queue-Powered Movie Discovery Platform · Feb 2026

A production-grade movie discovery platform powered by a self-updating database pipeline. Features a decoupled background worker for processing massive datasets and a high-performance, cinematic frontend.

React (Vite) Node.js MongoDB Redis BullMQ Framer Motion Tailwind CSS Render
Live Demo GitHub
  • Distributed Architecture: Designed a decoupled system with a FastAPI/Express REST API and a dedicated Background Worker running on Render to handle heavy data processing without blocking the UI.
  • ETL Data Pipeline: Engineered a Redis + BullMQ job queue to parse massive IMDb TSV datasets (millions of rows) and synchronize metadata with TMDB APIs in real-time.
  • Advanced UI Engineering: Built a polished interface using Framer Motion for 3D card tilts and implemented React Portals to solve complex mobile viewport clipping issues for modal overlays.
  • Performance & Caching: Integrated TanStack Query for server-state management, caching, and optimistic updates to ensure instant interactions.
  • Production Security: Implemented robust security practices including JWT authentication, dynamic CORS whitelisting for Vercel previews, and environment-based configuration.

Drowsiness Detection System Real-Time Computer Vision · Dec 2024

Real-time computer vision system that monitors eye state using facial landmark analysis to classify user attention as Active, Drowsy, or Sleeping, triggering alerts to prevent fatigue-related accidents.

Python Flask OpenCV dlib NumPy pygame React.js JavaScript Computer Vision EAR
GitHub
  • Implemented real-time eye-state classification using facial landmark detection (68-point model) and Eye Aspect Ratio (EAR) analysis to distinguish Active, Drowsy, and Sleeping states.
  • Built a Flask backend for live video processing and REST APIs, achieving ~92% detection accuracy with <100 ms frame processing latency at 30 FPS.
  • Designed alarm logic to trigger after 5 seconds of continuous eye closure, reducing false alerts by approximately 40% through blink filtering and temporal smoothing.
  • Developed a React-based dashboard displaying live video, real-time attention state, duration timers, and manual alarm controls.
  • Applied the system to real-world use cases including driver safety, workplace fatigue monitoring, and human–computer interaction research.

Riddhi’s Creation — Handcrafted Crochet Studio Content-Driven Brand Website · Jan 2026

Calm, premium digital lookbook for a handcrafted crochet brand, focused on storytelling, visual elegance, and CMS-driven content without e-commerce complexity.

Next.js (App Router) TypeScript Tailwind CSS GSAP Sanity CMS Vercel
GitHub
  • Designed and built a luxury, content-first web experience inspired by handcrafted artisan studios — intentionally avoiding carts, payments, and user accounts.
  • Implemented CMS-driven sections (About, Collections, Recent Orders) using Sanity with structured schemas and singleton documents.
  • Added cinematic GSAP scroll animations, full-screen hero sections, and mobile-first responsive layouts for immersive storytelling.
  • Integrated floating WhatsApp CTA to support direct inquiries while preserving a calm, non-commercial user experience.

SARIMA-Based Time Series Sales Forecasting System Time Series Analysis · May 2024

Category-wise time-series forecasting system using SARIMA models to predict monthly sales trends, capturing seasonality and long-term patterns for data-driven demand planning.

Python Pandas NumPy Statsmodels (SARIMAX) Matplotlib Scikit-learn Time Series Analysis SARIMA
GitHub
  • Built a category-wise sales forecasting pipeline using Seasonal ARIMA (SARIMA) models to analyze trends and monthly seasonality in historical supermarket sales data.
  • Implemented automated data preprocessing, monthly aggregation, stationarity checks, and SARIMA parameter tuning using ACF/PACF analysis.
  • Generated 12-month sales forecasts with confidence intervals, enabling clear interpretation of uncertainty and future demand.
  • Evaluated model performance using MAE, RMSE, MSE, and MAPE, achieving approximately 92% forecast accuracy and reducing RMSE by ~18% compared to baseline models.
  • Visualized historical vs forecasted sales trends with uncertainty bands to support inventory planning and strategic decision-making.

Experience & Training

Hands-on industry training and applied experience working with real-world datasets, analytics tools, and production-style workflows.

Data Analytics & Data Visualization Intern IBM SkillsBuild

Jun 2023 – Jul 2023
  • Completed an intensive Data Analytics and Data Visualization internship focused on end-to-end analytical workflows, including data cleaning, preprocessing, exploratory data analysis, and insight generation.
  • Analyzed real-world datasets using Excel, SQL, and Python (Pandas, NumPy) to identify patterns, trends, and anomalies, applying structured analytical methodologies.
  • Designed interactive dashboards and visual reports using IBM Cognos and Tableau to communicate KPIs and data-driven insights to stakeholders.
  • Strengthened data storytelling, statistical reasoning, and business-oriented decision-making by translating complex data into actionable recommendations.
  • Developed strong analytical thinking and problem-solving skills through hands-on exercises and case-based learning aligned with industry practices.
Python SQL Pandas NumPy Excel Tableau IBM Cognos Data Analytics Data Visualization

Education

Bachelor of Engineering in Computer Engineering

Vidyavardhini's College of Engineering and Technology | University of Mumbai 2021 – 2025

Focus areas included Artificial Intelligence, Machine Learning, Data Science, and Full-Stack Development.

Computer Engineering Artificial Intelligence Machine Learning Data Science Full-Stack Development

Let’s Connect

Open to full-time roles, internships, and meaningful collaborations. Feel free to reach out.

Email GitHub LinkedIn Resume

or email me directly at leanderfdes22@gmail.com