Ashwin K M ƒ +91-8838451653

Roll No.: 23882 # [email protected]

Bachelor of Technology # [email protected]
Mathematics and Computing § GitHub Profile
Indian Institute of Science , Bangalore ï LinkedIn Profile

About Me
I am a second-year B.Tech student in Mathematics and Computing at IISc, with a strong focus on integrating AI and
Machine Learning with risk analysis and quantitative finance. Passionate about advancing AI, I actively explore
Reinforcement Learning, Robotics Control, and Algorithmic Trading. Beyond academics, I enjoy reading fantasy novels,
staying updated on AI advancements, and engaging in debates and academic discussions. I also maintain discipline through
strength training and balance my rigorous coursework with gaming. My goal is to drive innovation in AI and ML, refining
my technical skills to make a meaningful impact in industry.

Education
• Indian Institute of Science , Bangalore 2023-2027
B.Tech in Mathematics and Computing CGPA : 7.2
• Sri Krish International School, Chennai 2021-2023
Central Board of Secondary Education Percentage: 90%
• Lalaji Memorial Omega International School 2009-2021
Central Board of Secondary Education Percentage: 98%

Projects
Implementation of Decision Transformers for Offline RL (Supervised by Prof. Chiranjib Bhattacharyya)
GitHub: Decision_Transformer_RL

(a) Implemented State-of-the-Art RL Models: Successfully reproduced Decision Transformers for offline RL, validating
its performance on benchmark datasets (D4RL). Implemented the model across locomotion tasks (HalfCheetah,
Hopper, Walker) and goal-conditioned environments (2D Reacher).
(b) Explored Credit Assignment in Long-Horizon Tasks: Investigated how transformers assign credit over long horizons
which traditionally required explicit state-associative reward modeling in RL.
(c) Leveraged Conditional Sequence Generation for Decision Making: Employed a trajectory representation including
returns-to-go, states, and actions, enabling autoregressive training and generation. Conditioned the model on target
returns to dynamically adjust reward conditioning and generate future actions
(d) Contributed to a Novel Paradigm in RL: Helped validate a paradigm shift where transformers—previously
underexplored in RL due to training challenges—demonstrated competitive performance against state-of-the-art
offline RL algorithms with minimal architectural modifications.

Advanced Data Structures and Algorithmic Implementations (Supervised by Prof. Pandurangan)

GitHub: UMC_201_Projects

(a) Ultra-Large ADT for Numerical Processing: Designed a high-performance system for ultra-large integer
computations. Developed liblargeinteger.so in C++ for optimized pair counting, achieving significant speedup.
(b) Optimized Dictionary with Tries: Replaced a naive dictionary ADT with a Trie-based implementation, eliminating
fixed-size constraints and enhancing memory efficiency and query speed.
(c) Efficient Query Handling with Splay Trees: Demonstrated splay trees as optimal for skewed query distributions,
achieving amortized O(1) access for frequently queried elements.
(d) Optimized Probabilistic Nearest Neighbor Search: Improved HyperLogLog’s nearest neighbor estimation by
replacing sorting-based selection with a faster heap-based approach, reducing computational overhead while
preserving accuracy.

Advanced Machine Learning: Deep Learning on Large-Scale Datasets(Supervised by Prof. Shishir Kolathaya)
GitHub: UMC_203_Projects

(a) Conducted a rigorous separability analysis on CIFAR-100, a state-of-the-art dataset, by implementing Perceptron
and Support Vector Machines (SVMs). Utilized both linear and Gaussian kernelized SVMs to diagnose and mitigate
non-separability issues.
 (b) Engineered and trained deep learning models including Multi-Layer Perceptrons (MLPs) and Convolutional Neural
Networks (CNNs) on a large-scale subset of MNIST-JPG. Achieved performance benchmarking using precision,
recall, and F1 score.
(c) Developed a Decision Tree Classifier on the UCI Heart Disease Dataset, optimizing hyperparameters via oracle
queries and identifying key cardiovascular risk factors through feature importance analysis.
(d) Developed Support Vector Regression (SVR) models for stock price prediction using historical stock market data,
leveraging slack linear and RBF-kernelized SVRs with multiple time window configurations.

Technical Skills
1. Computational Optimization: Algorithmic optimization, low-level memory management, cross-language integration
(Python & C++ using ctypes), and performance benchmarking for high-efficiency numerical computations.
2. Advanced Machine Learning & Optimization: Expertise in deep learning, SVMs, decision trees, and regression models
on industry-standard datasets; applied quadratic programming, feature engineering, dimensionality reduction, and
hyperparameter tuning to optimize model performance.
3. Languages : Proficient in Python, scikit-learn, Pandas, HuggingFace PyTorch, Tensorflow, Numpy, Matplotlib and
dtreeviz, leveraged visualization techniques and statistical metrics (precision, recall, F1-score, MSE, AUC-ROC) to
enhance model transparency and decision-making

4. Frameworks : Comfortable with Google Colab, LateX VS code, Conda, Jupyter Notebooks and Git

Positions of Responsibility
• Core Coordinator: Nexus, Rhapsody 2025, IISc Dec 2024 - March 2025
As the Core Coordinator of Nexus – SciTech Fest at Rhapsody, I led its inception, transforming it into a premier hub for
AI, Robotics, IoT, and Automation. I brought in top organizations like Autodesk, L&T, OLA, ArtPark, FSID, and
Incense, aligning their expertise with Nexus’ vision. I spearheaded Robonautica, Code Conquest, InnovateX, and
Intelligent Junkyard, IISc’s flagship competitions driving student innovation. Additionally, I established a platform for
startups to showcase their innovations, curated recruitment booths, and fostered industry partnerships, positioning Nexus
as a key driver of deep-tech collaboration and talent development

• Event Coordinator: RevCoding, Pravega 2025, IISc July 2024 - Jan 2025
Designed and organized RevCoding, IISc’s premier coding competition where participants deciphered hidden clues across
code-based challenges, integrating logical deduction with programming skills to solve complex problems.

• Sponsorship Coordinator: Pravega XI, IISc Aug 2024 - Jan 2025

As the Sponsorship Coordinator for Pravega XI, I spearheaded sponsorship acquisitions, securing substantial funding and
establishing long-term industry partnerships. I engaged with leading organizations, aligning their brand vision with
Pravega’s mission to drive scientific innovation. By fostering strong corporate relationships, I enhanced sponsor retention
and brand loyalty while expanding opportunities for student-industry collaboration. My efforts positioned Pravega as a
premier platform for academic excellence and deep-tech engagement.

Achievements
• World Scholars Cup Qualified at 30th Position in the Manilla Global Round 2019
The World Scholar’s Cup is an annual international academic program involving over 50,000 students from 60+ countries.
I competed in the Manila Global Round, achieving 8 Gold and 5 Silver medals and placing 30th out of 1,000 teams,
qualifying for Tournament of Champions held at Yale University

• NTSE Stage I Awardee 2021

Courses
Computer Science - Artificial Intelligence, Machine Learning, Automata Theory and Computability, Numerical Methods of
Optimisation, Data Structures and Algorithms, Discrete Mathematics, Computer Systems, Algorithms and Programming

Mathematics - Algebraic Structures, Linear Algebra, Real Analysis, Probability and Statistics
Breadth Soft Core Courses - Introductory Chemistry I(Physcial Chemistry), Introductory Biology I(Organismal Biology),
Introductory Chemistry II(Inorganic Chemistry), Digital Electronics, Analog Electronics, Introduction to Material
Sciences(Crystallography)