I am a Electrical Engineering undergraduate student at National Institute of Technology (NIT), Rourkela. I operate at the intersection of hardware systems, embedded engineering, and custom machine learning frameworks—focusing deeply on low-level execution, edge AI, and mathematical optimization over high-level abstractions.

"I am Atomic."

• Mathematical ML & Data Systems: Building machine learning algorithms (Gradient Descent, K-Means, KNN) completely from scratch to master optimization loops and algorithmic performance.
• Embedded AI & Automation: Designing embedded control systems (PID loops on Arduino) and deploying resource-constrained, compressed neural networks.
• Full-Stack Development: Architecting low-latency, highly interactive web applications using Next.js, Framer Motion, and Firebase for high-traffic student operations.

• Stack: Python, ONNX Runtime, OpenCV, NumPy, COCO Dataset
• Engineered a lightweight, real-time local edge inference engine running deep learning graphs optimized directly on the CPU.
• Eliminated heavy dependencies by utilizing NumPy for vectorized frame preprocessing (tensor reshaping, normalization) and writing a native Non-Maximum Suppression (NMS) parsing routine.
• Developed a dynamic live Heads-Up Display (HUD) to render bounding boxes, classifications, and real-time millisecond latency telemetry over a raw webcam feed.

• Stack: Pure C++, Python, Embedded Systems Optimization
• Built a zero-dependency, from-scratch C++ inference engine to deploy deep learning models onto severely resource-constrained edge hardware without relying on frameworks like TensorFlow Lite.
• Swapped out high-level abstractions for raw performance by manually implementing the entire forward pass, dot products, matrix multiplications, and non-linear activation functions ($\text{ReLU}$ and Linear) from scratch.
• Handled the entire development pipeline from Python-to-C++ serialization to final cross-compilation.

• Stack: Python, Deep Learning, Sequence Modeling
• Built an autoregressive deep learning model from the ground up designed to predict and suggest sequential text completions, mimicking an intelligent autocomplete engine.
• Curated and trained the network on a custom domain-specific dataset, achieving strong sequential coherence over deep text generation loops (e.g., iteratively resolving long-tail contextual strings like "mail once the course is over" from a single seed word).

• 💼 LinkedIn: linkedin.com/in/jatinbalajisailada/
• 📬 Email: jatinbalaji99@gmail.com