Pijus Saha
AI Engineer & Full-Stack Developer
B.Sc. CSE graduate specializing in deep learning, computer vision, and production full-stack engineering.
Technical Arsenal & Research Focus
Deep learning, computer vision, full-stack platforms, low-latency C++, and embedded IoT.
Deep Learning & Computer Vision
Neural Architectures, Vision Models & HTR
Hybrid CNN-BiLSTM-CTC architectures for handwriting recognition, multi-model benchmarking across CNNs and Vision Transformers, and image segmentation pipelines.
Full-Stack Web & Production Platforms
Modern Frameworks & Serverless Cloud
Performant web applications using Next.js 16 App Router, React 19, TypeScript, Tailwind CSS, shadcn/ui, Zod schema validation, and automated CI/CD on Vercel.
High-Performance C++ & Computer Graphics
2D/3D Rendering & Low-Latency Execution
Real-time 2D/3D graphics rendering in modern C++ and OpenGL with FreeGLUT, matrix math transformations, and algorithmic problem solving.
Scientific Python & System Automation
Desktop Tooling, File Orchestration & Analytics
Desktop GUI applications, collision-safe file orchestration, and progress analytics using Python, CustomTkinter, NumPy, Pandas, and Pillow.
IoT & Embedded Systems
Microcontrollers, Sensor Telemetry & Cloud Actuation
Hardware monitoring and control using ESP32 microcontrollers, ultrasonic telemetry, electrical noise suppression, and real-time cloud dashboards via Blynk IoT.
Projects & Research
Deep learning research, full-stack platforms, computer graphics, IoT, and desktop automation.
Handwritten Prescription Analysis: AI-Powered Medicine Recognition
Undergraduate Research Thesis Defense
End-to-end deep learning system to parse and recognize handwritten medical prescriptions using a hybrid CRNN architecture with bidirectional LSTMs and CTC loss.
Key Architectural & Production Highlights
3 of 6- Offline HTRRecognizes unstructured cursive medication scripts from static images using spatial-sequential modeling.
- 8-Model BenchmarkEvaluates standard CNN-RNN architectures, compound-scaled structures, and Vision Transformers.
- Deterministic Post-ProcessingCombines stochastic ML predictions with a rule-based database matching loop using Levenshtein Distance.
COSMOVERTEX Web Platform
Lead Full-Stack Developer · Official Educational Consultancy Platform
Production educational platform engineered with Next.js 16 App Router and React 19 Server Components, featuring automated dual-channel CRM lead processing, a lightbox gallery, and global destination portals.
Key Architectural & Production Highlights
3 of 5- Resilient Multi-Channel CRMEngineered dual-channel asynchronous lead dispatch integrating Resend Email API with real-time Google Sheets webhook automation and WhatsApp 1-click consultation.
- Accessible Lightbox ArchitectureBuilt a filterable media showcase with Framer Motion animations, keyboard trap navigation (Esc, Arrow keys), touch-swipe gestures, and zero-CLS image inspection.
- Modular Destination EngineDesigned a scalable routing architecture powering 8 international study portals with structured intake timelines, tuition matrices, and interactive UI tabs.
ByteSpace New — Frontend Technical Assessment
Technical Assessment Submission · Jr. Software Engineer (Frontend) at Doin Tech Limited
Pixel-accurate educational platform built with Next.js 16 and React 19, matching official Figma specifications for Doin Tech Limited.
Key Architectural & Production Highlights
3 of 6- Pixel-Accurate Landing & 3D HeroBlur-backdrop navbar, 3D floating decorations, and electric blue CTA banner matching Figma pixel-for-pixel.
- Course Catalog & Real-Time Filtering18-course catalog with debounced live search, multi-level filter popovers, and frosted-glass badge metrics.
- Dynamic Course Details & Reviews Suite3-tab layout (About, Lessons, Reviews) with instant tab switching, video modal preview, and sticky pricing card.
Pijus Saha — Personal Portfolio & AI Engineering Platform
Production Next.js 16 Platform · Autonomous AI Assistant · Interactive Design System
Modern, high-performance personal portfolio and interactive engineering platform built with Next.js 16 App Router, React 19, and Tailwind CSS v4, featuring an embedded autonomous AI assistant, bidirectional skill spotlight filtering, and zero-CLS theming.
Key Architectural & Production Highlights
3 of 5- Autonomous AI Portfolio AssistantBuilt-in conversational AI assistant powered by Google Gemini (with DeepSeek / Groq fallback mechanisms), grounded on structured project README knowledge bases and candidate telemetry.
- Interactive Skill Spotlight ArchitectureDynamic bidirectional filtering connecting floating skill badges with project cards, featuring animated glowing borders, synchronized counters, and instant smooth-scroll navigation.
- Production Full-Stack Next.js 16 & React 19Leverages React Server Components (RSC) for rapid initial page loads, structured JSON-LD semantic data, dynamic OpenGraph previews, and custom physics cursor.
Certifications & Competitive Programming
Codveda Technologies Machine Learning Internship, Harvard CS50P certification, DIU competitive programming finalist honors, and verified credentials.
Machine Learning Internship (Remote)
Codveda Technologies
CS50's Introduction to Programming with Python
Harvard University
Unlock the Algorithm Fall '23 - Finalist
Daffodil International University
DIU Take-Off Fall 2022 - Finalist
Daffodil International University
Education & Milestones
From secondary education through B.Sc. graduation honors and industry internship experience.
Secondary School Certificate (SSC) — Science
Science curriculum with strong foundations in Mathematics, Higher Mathematics, and ICT.
Academic Highlights & Distinctions
- Graduated with GPA 4.67 / 5.00
- A+ in Mathematics, Higher Mathematics & ICT
Higher Secondary Certificate (HSC) — Science
Higher secondary science curriculum with focus on ICT and analytic disciplines.
Academic Highlights & Distinctions
- Graduated with GPA 4.33 / 5.00
- A+ in Information & Communication Technology (ICT)
B.Sc. in Computer Science & Engineering
Undergraduate curriculum in Algorithms, Data Structures, Operating Systems, Computer Vision, and AI. Thesis on handwritten prescription recognition — perfect defense grade.
Academic Highlights & Distinctions
- Overall CGPA: 3.64 / 4.00
- Final 6-Credit Thesis Defense: 4.00 / 4.00
- Focus: Computer Vision, Deep Learning, Algorithmic Problem Solving
Machine Learning Intern (Remote)
Successfully completed a remote Machine Learning Internship at Codveda Technologies, demonstrating strong analytical thinking, data preprocessing pipelines, exploratory data analysis (EDA), and machine learning model development.
Key Responsibilities & Highlights
- 100% remote internship with structured weekly deliverables and milestones
- Exploratory Data Analysis (EDA) & preprocessing on structured data pipelines
- Model development, evaluation metrics & applied machine learning algorithms in Python
- Awarded official Internship Completion Certificate with organizational commendation
Contact & Collaboration
Open for applied AI engineering roles, research collaborations, full-stack development, and graduate study opportunities.
Direct Contact Channels
Send a Direct Message
Project inquiry, research collaboration, or opportunity? Send a message.