>joining aws rds · aug 2026
Ashish Khadka
// distributed systems · databases · infrastructure
Software engineer specializing in distributed systems, databases, and infrastructure. Currently finishing my MS at University of New Haven — joining AWS RDS in August 2026.
01 about
Software engineer specializing in distributed systems, databases, and infrastructure.
Grew up in Nepal, studied at Deakin University in Melbourne, and currently finishing my MS in Computer Science at the University of New Haven. Along the way: ML anomaly detection research, production backend systems in Kathmandu, and a summer at Amazon. In August 2026 I'm joining AWS RDS — working on the infrastructure that powers databases at planetary scale.
Outside of systems work: I run, I compete in speech contests (Toastmasters International finalist), and I'm building DWAL — a distributed write-ahead log service with quorum-based durability, total ordering via monotonically increasing LSNs, and crash recovery.
location
Connecticut, US
starting
AWS RDS · aug 2026
education
MS CS · UNH
02 experience
Joining the Amazon RDS team to work on database infrastructure at scale.
Built internal tooling and services on the AWS platform. Details under NDA — happy to discuss in interviews.
Provost-funded research on LLM lateral thinking benchmarking. Published at IEEE FLLM 2025.
Led backend development of a high-performance server-side application. Architected PostgreSQL and MongoDB storage, implemented Docker/Kubernetes CI/CD pipelines, and established API security practices around OAuth and JWT.
Researched deep semi-supervised ML for anomaly detection using TensorFlow and PyTorch on MNIST and CIFAR-10. Concurrently designed the Faculty Learning Innovations website and built Tableau dashboards for the university LMS.
03 projects
A fault-tolerant, distributed logging service built for crash recovery and replication. Designed to stay consistent under node failures, network partitions, and partial writes.
$ more projects building...
github.com/khadkaashish →04 research
Are Language Models Good at Lateral Thinking?
2025 3rd International Conference on Foundation and Large Language Models (FLLM) · IEEE
Assessed LLMs on lateral thinking puzzles using NYT Connections and LinkedIn Pinpoint. Benchmarked four distilled models — LLaMA-3.1-8B-Instruct performed best with 31.50% on Pinpoint and 19.33% on Connections. Proposed first cybersecurity applications of lateral thinking AI.
Ashish Khadka, Mohamad Nassar, Shivanjali Khare
Lateral Thinking in LLMs: Benchmarking and Fine-Tuning for Puzzle Solving and Generation
Master's Thesis · University of New Haven
Extended the lateral thinking benchmark to include fine-tuning experiments. Studied how LLMs can be adapted to not just solve but generate lateral thinking puzzles. Explored the gap between reasoning ability and creative divergent thinking in modern language models.
University of New Haven
Deep Semi-Supervised Anomaly Detection (DSAD)
Research Study · Deakin University
Explored deep semi-supervised ML algorithms for anomaly detection. Trained and evaluated models using TensorFlow and PyTorch on MNIST and CIFAR-10 datasets, optimizing detection accuracy in low-label settings.
Professors and PhD researchers, Deakin University
05 skills
06 writing
Why is React JS so popular in 2023?
I am writing this blog for people who want to start building websites and mobile apps but don’t know which language to start with. I starte…
How to develop a habit of running
“Running is the greatest metaphor for life because you get out of it what you put into it.” — Oprah Winfrey, media executive, actress, and …
Melbourne 10k run ️
“It’s very hard in the beginning to understand that the whole idea is not to beat other runners. Eventually, you learn that the competition…
07 contact
get in touch
Distributed systems, research, opportunities, or just engineering in general — my inbox is open.