Bhavesh Goel

I've built AI-enabled web applications, backend services, and data pipelines across AceIPM, KaggleX Fellowship, and PropReturns. My work spans live examination platforms, OCR workflows, LLM fine-tuning, semantic search, and automated data ingestion.

At AceIPM, I developed and maintained a Next.js and Django computer-based testing platform and optimized REST scoring services handling 30K+ submissions on peak exam days. I also created a Word-to-LaTeX parser and ETL pipeline for 10K+ questions, automated ingestion for 700+ scanned PDFs, and delivered Generative AI-assisted teaching tools.

During the KaggleX Fellowship, I fine-tuned Gemma 2 2B for ICD-10 medical coding using PEFT/LoRA, achieving 92.18% sparse categorical accuracy and earning a $1,000 scholarship for model optimization and training-efficiency work.

Previously at PropReturns, I built Selenium and Pandas pipelines processing 200K+ housing records into AWS S3 and PostgreSQL, while an NLP extraction and translation pipeline reduced third-party translation spend by more than INR 5 lakh. I enjoy building practical systems from ingestion through deployment using Python, FastAPI, Django, React, PostgreSQL, and Docker.