Rhythm Bhetwal

1 Experience

Where the work has taken me

Proof of Work · Technical Signals

github.com/mythrhyth

Active Builder

Consistently shipping production AI systems, full-stack open source tools, and academic research code.

37+ Public Repos 9 Shipped Projects Open Source

GitHub Commit Activity (Heatmap) Live Sync

Rhythm Bhetwal GitHub Contribution Graph

Target Palette: #8a5cd6 User: @mythrhyth Feed: Real-time Activity

leetcode.com/u/mythrhyth

Active Problem Solver

Actively solving complex algorithmic problems across multiple domains, including dynamic programming, greedy methods, trees, and backtracking.

Competitive Programming Advanced DSA C++ · Python

LeetCode Solved Breakdown & Ranking Live Stats

Rhythm Bhetwal LeetCode Stats and Breakdown Card

Profile: @mythrhyth Focus: Advanced DSA Global Rating: ~1600

Sep ’26 – Present

Software Engineer now

Startup Crafters LLC

Building and shipping AI Agentic SaaS products with autonomous tool calling and structured workflows.
Designing full-stack interfaces, modular REST APIs, and asynchronous execution pipelines.
Driving production feature releases, performance optimizations, and scalable deployment stability.

Python TypeScript AI Agents FastAPI SaaS Architecture Docker

Jul – Sep ’26

AI Researcher

NAAMII · CESP Lab

Conducted medical AI research on multimodal endoscopy video and high-resolution imaging datasets.
Studied and evaluated deep learning architectures for clinical grading standards (MES and UCEIS severity scoring).
Executed GPU-accelerated model experimentation, literature reviews, and benchmark evaluations across AUROC, F1-score, and accuracy.

PyTorch Medical AI Computer Vision Multimodal Learning Deep Learning GPU Research

Jun – Sep ’26

Software QA / QA Engineer

Sentrinus

Designed and executed 1,000+ test cases across cybersecurity SaaS workflows covering functional, regression, and edge paths.
Identified and documented 50+ bugs with reproducible steps, system logs, severity context, and root-cause observations.
Contributed Vitest unit test suites for utilities and custom React hooks across 40+ codebase files.
Reduced testing false positives by approximately 50% through improved validation and test-to-requirement traceability.

TypeScript JavaScript Vitest Regression Testing Functional QA Cybersecurity SaaS

May – Dec ’25

AI/ML Fellow

Fusemachines

Built a reasoning LLM Hallucination Detection System utilizing a recursive hybrid LSTM + DeBERTa-v3-small architecture.
Achieved ~87% accuracy on custom evaluation benchmarks for reasoning-oriented outputs with sequences up to 256.
Constructed end-to-end evaluation and benchmarking pipelines with MLflow for experiment tracking and metric analysis.
Containerized model services with Docker and deployed low-latency inference endpoints via FastAPI.

Python PyTorch Transformers DeBERTa LSTM MLflow Docker FastAPI

2 Research

What I'm reading and writing

Targeting IEEE TBME

RA-CNN for ECG Arrhythmia Classification

A Residual Attention CNN evaluated under the stricter inter-patient AAMI DS1/DS2 protocol on MIT-BIH data.

CESP Lab · NAAMII

UCEISNet — Ulcerative Colitis Severity Scoring

Colonoscopy-based severity scoring on the EndoUC dataset, combining component-level UCEIS decomposition with direct scoring and caption generation.

Targeting Springer

Transformer-based Human Activity Recognition

IMU-based activity recognition reaching 99% accuracy on the VTT-ConIot dataset.

NITM AI Lab

WMT26 Low-Resource Translation

Contributed to a shared-task paper on English↔Arabic and Arabic↔Hindi translation using NLLB and mBART.

3 Recognition

Wins along the way

Winner

QueueCare ’26 Hackathon

Real-time queue management platform for live updates, designed accessibility-first.

Runner Up

Expiry Alert Hackathon

Industries track — enterprise concept to track product expiry and cut waste.

Winner

National Ideathon

Pitched a long-term ICT solution spanning 8 key sectors of the country.

Selected

LinkedIn CoachIn Mentee

Chosen from 21,000+ applicants for a 6-month 1:1 mentorship with LinkedIn engineers.

Fusemachines AI micro-degree certificate

Fellowship

Fusemachines AI Fellow

Six-month AI fellowship — capstone: an LLM hallucination-detection pipeline (87% accuracy) and the OmniRAG system.

4 Community

Giving back

I like mentoring people, sharing what I know, and a good coffee chat about tech.

Campus Mantri

GeeksforGeeks

Promoting GfG products through campaigns, workshops, and tutorial sessions.

Student Ambassador

Microsoft Learn

Representing Microsoft and its products through workshops in my community.

AI/ML Coordinator

Coding Club, NITM

Organizing workshops, coding contests, and hackathons for the student community.

6 Selected Work

Built & shipped

NITMedi

Medical inventory & insurance tracking for campus. Live on the intranet, used by 1,000+ students daily.

Case study

TypeScript

QueueCare

Event-driven hospital queue system with a wait-time prediction engine trained on consultation history.

Case study

Co-built

OmniRAG

Retrieval pipeline across text, image, and audio queries — every answer traces back to its source chunk.

Case study

Python

Stock Market Prediction

End-to-end forecasting pipeline on live market data with an interactive Streamlit dashboard.

Case study

TypeScript

NITM MessTrack

Official mess management system for NITM — meals and dues, tracked for the whole hostel.

Case study

Logistics

LoadFlow

Real-time brokerage platform sharing live data between brokers, dispatchers, and sellers.

Case study

Enterprise

Expiry Alert

Management system that tracks product expiry across a supply chain to reduce waste.

Case study

Deep Learning

Brain Tumor Classifier

96% accuracy among 100+ competing teams on MRI scan classification.

Case study

NLP

LSTM Story Generator

A sequence model that writes short stories from a starting prompt.

Case study

7 Certifications

Courses and credentials

Coursera

Supervised Machine Learning

Regression and classification — DeepLearning.AI & Stanford Online, Oct 2024.

Coursera

Advanced Learning Algorithms

Neural networks and decision trees — DeepLearning.AI & Stanford Online, Oct 2024.

Unsupervised Learning, Recommenders, Reinforcement Learning certificate

Coursera

Unsupervised Learning & RL

Recommenders and reinforcement learning — DeepLearning.AI & Stanford Online, Oct 2024.

PDSC

Understanding LLMs and RAG

PDSC power series — Institute of Engineering, Pulchowk Campus.

LOCUS 2025

Hardware Fellowship

8-day fellowship — Robotics Club, IOE Pulchowk, Tribhuvan University.

Hackathon

HackVega 2.0

Nationwide hackathon — competed against 47,000+ engineering students.

McKinsey & Company

Forward Fellow

Completed the McKinsey Forward program focused on foundational leadership and business skills.

Rhythm
Bhetwal.

Where the work has taken me