About Me

Introduction

My name is Harsha, currently pursuing an MSc in Computer Science at Kristu Jayanti University in Bangalore. My current focus is on system design, Python and FastAPI, with most of my work moving deeper into AI and backend engineering.

Some of my recent work has been around RAG architectures and agentic tooling. Some of it is academic in nature, the rest lives on my GitHub.

I am also open to AI engineering internships and entry-level roles, particularly those involving applied LLMs or backend infrastructure. page.

Work

Although I was introduced to software through front-end development, I have moved steadily toward backend systems and applied AI. The closer I get to data, infrastructure and the systems behind the model, the more interesting the work becomes. I enjoy thinking about system design, evaluating tradeoffs empirically, and building tools that solve a real problem rather than demonstrate one.

Most of my recent work has been in Python and FastAPI, across both general backend services and AI-adjacent systems. I use JavaScript and React when a project needs a front-end, though this is the exception rather than the rule in what I build now.

Career

I started programming early - curiosity, a slow computer, and a need to understand how things worked. Somewhere in the middle I took a detour into biology. It didn’t take. Computing pulled me back, and this time it stuck.

I began with small front-end experiments and gradually moved into full-stack work, and then into applied AI systems. My recent projects reflect this trajectory:

  • Scholar, a full-stack RAG application with a hybrid retrieval pipeline (BM25, vector search, Reciprocal Rank Fusion, Cohere reranking), paired with an empirical study of 21 chunking-and-retrieval combinations analyzed with two-way ANOVA;
  • Orion, a Company Intelligence Agent built with FastAPI and LLM tool-calling; and more recently
  • Transcriber, a lightweight Electron desktop app that turns local video files into plain-text transcripts entirely offline, using faster-whisper running on CPU and ffmpeg for audio extraction - vibe-coded with Claude Code (Sonnet 4.6), more an exercise in shipping than in careful design.

Beyond the technical, the bigger shift over the last few years was internal. Discipline replaced talent. Systems replaced motivation. Running a full marathon in October 2024 reshaped how I think about long effort — distance doesn’t negotiate, you train, you suffer, you finish. Debugging at 2 a.m. feels familiar for the same reason.

Education

I am currently pursuing an MSc in Computer Science at Kristu Jayanti University in Bangalore, in the final stages of the programme.

Before this, I completed a BSc (Honours with Research) in Data Sciences and Analytics at MS Ramaiah University of Applied Sciences, with a CGPA of 8.17.

Alongside formal study, I spend a fair amount of time on self-directed learning. At the moment that means working through Python at the level of internals, data structures and the standard library, because I believe that to write software properly you need to understand the tools beneath what most people take for granted. Learning this way has changed how I approach writing software and has given me a deeper appreciation of the craft.

Certificates

In addition to formal education, I completed the Full Stack Development Path at Scrimba.

Leisure

In my spare time I read a lot, mostly non-fiction — currently The Courage to be Disliked by Ichiro Kishimi and Fumitake Koga. My personal interests sit around philosophy, history and the way ideas travel between disciplines.

Apart from reading, I write poetry and reflective essays, watch and dissect films, and play chess occasionally. I also play darts, where accuracy matters more than force, and I run — the marathon I completed in October 2024 being the most demanding endurance project I’ve taken on so far.

Website

This website is intentionally simple. No JavaScript. No trackers. No analytics. It is built to load fast, age well, and keep the focus on words and ideas. Posts are written in Markdown.

The source code is available on my GitHub. This site is not a highlight reel — it is a record of work, thought and progress.