π§ chinmayjain@iisc.ac.in
π π GitHub | π LinkedIn
Hi! Iβm Chinmay Jain, a passionate researcher and engineer currently pursuing my M.Tech in Computer Science and Engineering at the Indian Institute of Science (IISc), Bangalore. With a strong foundation in Electrical Engineering and professional experience at Deloitte India, I have cultivated a unique blend of analytical thinking and hands-on technical expertise.
My academic journey has led me to specialize in Machine Learning, Natural Language Processing (NLP), and High-Performance Computing, where I explore innovative solutions to complex problems. I am particularly focused on optimizing recommendation systems by eliminating traditional embedding tables and leveraging advanced hashing techniques to achieve significant memory and computational efficiency without compromising accuracy.
Iβm deeply motivated by the challenge of bridging the gap between theoretical insights and real-world applications. Whether it's developing models to improve linguistic accessibility, enhancing the computational performance of AI systems, or building autonomous robotics solutions, I strive to make a meaningful impact in the field of technology.
When Iβm not immersed in research or coding, youβll often find me sharpening my skills in chess, hitting the badminton court, or enjoying anime. I believe in the power of interdisciplinary growth and the importance of combining creativity with technical rigor to solve problems effectively.
- Accelerating Recommendation Systems (Thesis): Optimized inference by replacing traditional embedding tables with hash embeddings, reducing memory usage by 80%.
- Dilated Convolution Optimization: Enhanced performance through multi-threading, CUDA GPU implementation, and SIMD vectorization.
- Hardware Performance Profiling: Analyzed branch predictors (GShare, Perceptron, TAGE) on SPEC benchmarks, achieving up to 98.5% accuracy in hybrid configurations.
- Low-Resource Language Translation: Fine-tuned state-of-the-art LLMs like LLaMA and Mistral, improving BLEU scores by 4 points for Hindi-English and English-Assamese translations using PEFT and LoRA.
- Seq2Seq Transliteration of Indian Names: Built an RNN-based model with attention mechanism for transliterating names into Hindi, achieving BLEU: 0.8569 and reducing token error rates.
- Sentiment Classification: Created models using rule-based classifiers and Word2Vec-based architectures, achieving up to 72% accuracy.
- Fine-Tuning GPT using LoRA: Efficiently fine-tuned GPT-2 for linguistic acceptability, reducing trainable parameters by 99.5% and achieving 85.83% accuracy.
- Knowledge Distillation: Transferred knowledge from GPT-2 to a smaller RNN model, achieving 79.26% accuracy on CoLA dataset.
- Image Classification and Contrastive Learning: Implemented softmax regression and CNNs on CIFAR-10, achieving 82.38% validation accuracy with contrastive learning techniques.
- Community Detection: Applied Girvan-Newman and Louvain algorithms to real-world datasets, uncovering modularity-based communities effectively.
- Three-Wheeled Holonomic Drive System: Built a robotic platform for motion planning using deduced reckoning localization.
- Autonomous Quadruped Jumping Robot: Designed a four-legged robot with adaptive gaits and real-time sensing using LIDAR and gyroscopes.
- M.Tech, Computer Science and Engineering | IISc Bangalore | 2023β2025
- B.Tech, Electrical Engineering | Nirma University | 2018β2022
Programming Languages: C, C++, Python, SQL Libraries & Tools: PyTorch, NumPy, TensorFlow, Scikit-Learn, Pandas, git, pandas ML/DL: Transformers, CNNs, RNNs, Contrastive Learning, Knowledge Distillation , PCA, SVM, LSTM/GRU, BERT, GPT, RAG.
- Natural Language Processing
- Machine Learning
- High-Performance Computer Architecture
- Probability And Statistics
- Systems for Machine Learning
- Data Analytics
- Design And Analysis of Algorithms
- Applied Linear Algebra And Optimisation
-
Analyst, Deloitte India
Streamlined data migration processes and optimized reporting systems using SAP-ERP and ABAP tools. -
Intern, Deloitte India
Developed front-end CRUD applications with Fiori/UI5 and enhanced data navigation interfaces.
I am always eager to explore opportunities in machine learning, natural language processing, and high-performance computing. Feel free to reach out for collaborations, research, or brainstorming innovative ideas!