I have 2+ years of experience in data and developer roles across the marketing and finance sectors through internships and co-ops. These experiences pushed me to build robust, scalable solutions that bridge data, product thinking, and engineering. I constantly seek new challenges and opportunities to expand my expertise.
This portfolio showcases selected projects across various domains: LLMs, NLP, CV, forecasting, analytics, and backend development.
| Tag | Project | Description |
|---|---|---|
| LLM | Fast Inference of LLMs via Speculative Decoding | Implemented Speculative Decoding and BiLD to accelerate Transformer inference. Benchmarked across models. |
| LLM | GPT-2 from Scratch (ongoing) | Reproducing GPT-2 124M model from scratch to understand Transformer internals and training loops. |
| LLM / NLP | ChronoEvents: News Timeline Builder | RAG system that transforms podcast news transcripts into summarized event timelines. |
| LLM / Backend | MCP Server with Claude | Server tool for real-time document retrieval via Claude and the Serper API. |
| NLP/ML | Multi-Label Content Categorization System | Multi-label news categorization using NLP and ML - classify news articles into 17 categories with a live demo and REST API. |
| Backend (JavaScript) | Subscription Tracker | Built a backend system with JWT authentication, DB modeling, API architecture, security and automated workflows |
| Audio / CV | Audio Deepfake Detection with LCNN | Built a deepfake audio detection system using LCNN with self-attentive pooling. |
| CV | Scene Text Recognition for Jersey Numbers | Two-stage OCR pipeline to extract jersey numbers from blurry, occluded sports footage. |
| ML | Rohlik Orders Forecasting | Multi-warehouse time series model to forecast 60-day order volume for e-grocery operations. |
| ML | Rossmann Sales Data Prediction | Forecasted daily retail sales across 1000+ stores using time series regression models. |