Yadunandan
M Nimbalkar
AI/ML Engineer in the Making
Global Academy of Technology · 2023 – 2027 (Expected)
Actively looking for AI/ML internships and collaborative projects.
Get in touch →Curious by nature,
precise by practice.
I'm a Computer Science Engineering student specialising in Artificial Intelligence and Machine Learning at Global Academy of Technology, Bengaluru. Fascinated by how systems work internally, I build projects that sit close to the implementation layer — understanding models, data pipelines, and real-world deployment.
My focus is on the implementation layer — understanding how models work internally, how data pipelines are structured, and how systems get deployed in real environments rather than just using high-level abstractions.
Currently pursuing my BE in CSE (AI & ML) at Global Academy of Technology, maintaining a CGPA of 8.8.
Coding Curiosity Begins
Started exploring programming and technology fundamentals at Sri Vidya Kendra.
Structured Academics
Grade 10 CBSE. Built strong mathematical and analytical foundations.
Higher Secondary & NCC
Completed PCMC (82%) at Podar International School. Earned NCC A-Certificate.
Joined GAT · AI/ML Track
Started BE in CSE (AI & ML) at Global Academy of Technology, Bengaluru.
First ML Projects + IBMZ 2024
Built Toxic Plant Identification (CNN) for IBMZ Datathon 2024. Explored voice ML with Parkinson's Analyser.
IBMZ 2025 & RAG Systems
Built ASD Early Screen at IBMZ 2025. Built AI Codebase Analyzer with FAISS + RAG pipeline.
Production Systems
Building a production Recommendation System. Targeting AI/ML internships.
Skills & Proficiency
Honest self-assessed proficiency levels across languages, ML, frameworks, and CS fundamentals.
Languages
AI & ML
Frameworks
Fundamentals
Projects
5 completed projects spanning ML systems, NLP, Computer Vision, RAG pipelines & healthcare AI — including 2 IBM-Z Datathon builds.
Two-Tower Recommendation System
Production-grade personalized movie recommendation engine. Maps users and items into a shared 128-dim embedding space via a Two-Tower neural model (InfoNCE loss + in-batch negatives), then re-ranks candidates with a GBM model achieving Val AUC 0.9799. FAISS IVF index delivers 4.5× faster retrieval at 99.8% recall. LRU cache hits ~55% within session with 0 ms cached latency. Median API response ~19 ms.
- ▸25M ratings · MovieLens
- ▸Recall@100 = 15.7%
- ▸GBM Val AUC = 0.9799
- ▸FAISS 4.5× faster retrieval
- ▸~19 ms median API latency
AI Codebase Analyzer
LLM-powered developer tool that analyzes any public GitHub repo via a RAG pipeline. Clones repo, chunks source files, indexes embeddings in FAISS, and answers natural language questions grounded in actual code. Includes AST-based dependency graph for module-level import analysis.
ASD Early-Screen
Risk-screening tool predicting likelihood of Autism Spectrum Disorder from behavioral/developmental responses. Features AES-128 artifact encryption (model encrypted at rest, decrypted in-memory only), SHAP explainability per prediction, SHA-256 hashed audit logs with no PII, and Flask web UI deployed on IBM LinuxONE.
- ▸~390 ms inference + SHAP
- ▸Live on IBM LinuxONE
- ▸AES-128 model encryption
Parkinson's Voice Detection
Detects Parkinson's Disease from 3 seconds of sustained vocal phonation with no language constraint. Solo v2 rebuild — fixed 6 critical bugs in the original codebase (aliased Praat calls, silently null nonlinear features, subject data leakage). 56-feature acoustic pipeline + 75 Optuna trials. wav2vec2-XLS-R-300M embeddings for cross-lingual generalization.
- ▸CV AUC = 0.972 ± 0.034
- ▸CV Accuracy = 94.2%
- ▸Subject AUC = 0.996
- ▸831 recordings · 61 subjects
Toxic Plant Identification
Binary CNN classifier identifying toxic vs non-toxic plants from photo uploads. Built image preprocessing pipeline with OpenCV, trained on Kaggle dataset, and deployed as a Flask web app with real-time prediction interface.
Achievements & Certifications
Recognition and learning credentials collected along the journey.
Competition · Code Warriors GAT-174
Built ASD Early Screen — a secure, explainable ML tool for early autism risk detection with SHAP explainability and AES-encrypted artifacts, deployed on IBM LinuxONE.
Competition Participant
Built Toxic Plant Identification — a CNN-based binary classifier (toxic/non-toxic) with a Flask web interface for real-time image-based predictions.
National Cadet Corps
Earned the NCC A-Certificate, demonstrating discipline, leadership, and commitment to national service.
Certification
Completed certification covering Generative AI fundamentals — LLMs, diffusion models, prompt engineering, and real-world applications.
Certification
Certified in Large Language Model fundamentals including transformer architecture, fine-tuning strategies, and deployment patterns.
Certification
Foundational certification covering AI problem framing, model selection, evaluation, and practical implementation.
Certification
Comprehensive certification on computer networking fundamentals — protocols, OSI model, TCP/IP, and network architecture.
BE in CSE (Artificial Intelligence & Machine Learning)
Global Academy of Technology, Bengaluru
CGPA: 8.8 / 10 · 2023 – 2027 (Expected)
Coursework: Design & Analysis of Algorithms, Data Structures, Machine Learning, Database Systems, Operating Systems, Artificial Intelligence
Let's Connect
Open to internships, collaborations, and conversations about AI/ML. If you're working on something interesting, reach out.
Designed & built by Yadunandan M Nimbalkar · 2025