Manthan Mittal
AI Engineer • Full‑Stack Developer • Creator
Architecting intelligent systems and scalable applications. Specializing in Generative AI, LLMs, and Full-Stack Engineering.
About Me
I exist at the intersection of scalable engineering and bleeding-edge AI. As a Full-Stack AI Engineer, I don’t just train models—I build autonomous systems that solve hard problems. My technical philosophy is grounded in the belief that true innovation happens only when robust software engineering practices are fused with advanced artificial intelligence.
My background is rigorous and diverse, spanning from optimizing mobile architectures for speed to researching complex Deep Learning systems during my M.Sc. in the UK. This academic focus culminated in my thesis on "Random MAC Optimization," where I investigated novel strategies to push the boundaries of algorithmic efficiency and model convergence in resource-constrained environments.
Beyond theory, my work is backed by data and community impact. I have developed open-source tools used by thousands of developers and architected production-grade RAG pipelines. Whether it involves deploying large-scale web agents or optimizing neural networks for robotics, I am dedicated to building software that thinks.
NEURAL NETWORKS
Technologies and domains powering intelligent systems
LANGUAGES
FRAMEWORKS
TOOLS
DOMAINS
CHRONICLES
EXPERIENCE
AI Research Intern
- ▹Engineered a Text-to-Video generation prototype leveraging Stable Diffusion and PyTorch, enabling high-fidelity video synthesis from natural language prompts.
- ▹Designed and implemented a custom CLIP-based similarity scoring pipeline to quantitatively evaluate semantic alignment between input prompts and generated video frames.
- ▹Authored a technical retrospective on generative AI challenges; the article was featured on the company blog and garnered 5,000+ reads.
Android Developer Intern
- ▹Developed scalable mobile applications using Flutter and Kotlin, implementing Clean Architecture principles to ensure code maintainability.
- ▹Integrated RESTful APIs and Firebase Authentication to support real-time data synchronization and secure user management.
- ▹Optimized application performance by implementing lazy loading and code-splitting, resulting in a 35% reduction in page-load time.
Software Engineer (Freelance)
- ▹Web Task Autopilot: Architected end-to-end web automation agents using Selenium and Python to streamline repetitive browser-based workflows.
- ▹DocInsight (RAG System): Built a Retrieval-Augmented Generation Q&A system using LangChain and OpenAI to allow users to query complex documentation naturally.
- ▹Open Source Leadership: Maintained multiple open-source libraries, achieving a cumulative 2.5k+ stars on GitHub and fostering a community of active contributors.
EDUCATION
M.Sc. in Artificial Intelligence
- ▹Specializations: Deep Learning, Robotics, Computer Vision, Natural Language Processing (NLP), and Machine Learning.
- ▹Master’s Thesis: “Random MAC Optimization” – Investigated optimization algorithms to enhance model convergence and efficiency.
B.E. in Information Technology
- ▹Graduated with First Class Honors.
- ▹President of the Coding Club – organized hackathons and workshops.
INNOVATIONS
Deploying intelligence into production.
AI Video Generator
Converts text prompts into short AI-generated videos using diffusion and transformer pipelines.
- •Implemented latent diffusion models for high-fidelity generation.
- •Optimized inference pipeline for faster video synthesis.
Intelligent Document Q&A System
Upload documents and get accurate answers using embeddings and a retrieval-augmented generation pipeline.
- •Built a scalable RAG pipeline for handling large PDF documents.
- •Integrated vector search for semantic query understanding.
Autonomous Web Navigator
An AI agent that performs tasks in a browser using LLM reasoning + Playwright.
- •Designed an agentic loop for multi-step web interactions.
- •Achieved high success rate on complex form-filling tasks.
Sleep Pattern Analyzer
Analyzes sleep data to detect irregularities using ML classification models.
- •Trained classification models to detect sleep anomalies.
- •Visualized sleep metrics with interactive Streamlit dashboards.
Random MAC Optimization
Uses neural networks & graph neural networks to optimize MAC layer scheduling.
- •Applied GNNs to model wireless network topology.
- •Improved scheduling efficiency compared to traditional algorithms.
Roxigym – Full-Stack Fitness Web App
Fitness app with login, workout routines, progress tracking, exercise library, and nutrition logging.
- •Developed full-stack features including auth and database management.
- •Deployed a responsive PWA for mobile users.
Autonomous Hybrid Navigation System using ROS2
A hybrid autonomous navigation framework for mobile robots that intelligently switches between PID-based wall following and Fuzzy Logic obstacle avoidance based on real-time LIDAR sensor data.
- •Implemented PID Controller for precise wall following and edge tracking.
- •Developed Fuzzy Logic Controller for smooth obstacle avoidance.
- •Designed Hybrid Selector to dynamically switch strategies based on LIDAR data.
Rossmann Store Sales Forecasting
Forecasted daily sales for 1,115 stores using a hybrid approach of Deep Learning (LSTM, RNN) and Ensemble Methods (XGBoost, Random Forest).
- •Engineered complex temporal features and handled seasonality to reduce RMSPE.
- •Conducted comparative analysis of Entity Embeddings in Neural Networks.
- •Combined Deep Learning and Ensemble Methods for robust forecasting.
Feedforward Neural Network Implementation for Robotic Kinematics Prediction
Implemented a Feedforward Neural Network to predict robotic arm kinematics, mapping joint angles to end-effector positions.
- •Trained FNN to solve forward kinematics problems with high accuracy.
- •Optimized model architecture for real-time inference in robotic control loops.
- •Analyzed prediction errors across the robot's workspace.
ESTABLISH UPLINK
Initiate communication protocol. Open to collaboration on AI research and full-stack development.