Experience
Saksham Rathi | CSE, IIT Bombay
Key Projects
All of my projects can be found on my GitHub profile. Below are brief descriptions of selected projects:-
Analyzing Selfish Mining and Eclipse Attack in a P2P Cryptocurrency Network
Guide: Prof. Vinay Ribeiro | Course Project: Blockchains, Cryptocurrencies and Smart Contracts
- Developed a P2P blockchain simulator with 100+ nodes, PoW mining and fork resolution under delays
- Simulated Selfish Mining + Eclipse Attacks; showed >50% malicious nodes cause full chain takeover
- Proposed a trust-based defence reducing malicious block acceptance by 95% with 40% malicious actors
-
C Compiler Development
Guide: Prof. Uday Khedker | Course Project: Implementation of Programming Languages
- Built a simplified C compiler in C++ using Lex/Yacc with lexer, parser and context-free grammar
- Engineered end-to-end compilation pipeline including abstract syntax tree construction, intermediate representation, basic optimizations and MIPS code generation in six incremental stages
- Integrated robust error handling in all phases to detect and reject syntactic and semantic violations
-
Automated Index Creation in a Database
Guide: Prof. Sudarshan | Course Project: Database and Information Systems
- Built an automated PostgreSQL indexer that adapts to query patterns, reducing execution time by 40%
- Used attribute frequencies and planner feedback for adaptive indexing without DBA intervention
- Designed a live query monitor with cost-aware filtering via hypopg and custom tracking structures
-
Denoising Diffusion Probabilistic Models
Guide: Prof. Sunita Sarawagi | Course Project: Advanced Machine Learning
- Implemented and trained conditional and unconditional DDPMs with UNet, experimenting with linear/cosine/quadratic schedules and achieved KL = 0.30, EMD = 9.65 on 32k-sample Albatross dataset
- Applied Classifier-Free Guidance and used conditional DDPM as a classifier (95% on 2-class tasks)
- Used SVDD for reward-guided sampling from unconditional DDPMs using classifier-derived reward signals
-
Guide: Prof. Ashutosh Gupta | Course Project: Data Structures and Algorithms
- Built trading strategies using sockets and threads to improve execution speed and market response
- Developed a dynamic market matching engine with arbitrage detection and prevention mechanisms
- Implemented median trading and statistical arbitrage to exploit price anomalies for steady returns