I don’t even know why I have a portfolio website at this point.
I work at the Centre for Data Science and AI at the University of Glasgow, where I also did my MSc. This is where I keep the stuff I’ve made.
A bit about me.
I did my engineering degree in AI and data science in Karnataka, then moved to Glasgow for my MSc and stayed. These days I work at the university’s Centre for Data Science and AI, on the admin side.
Before this, most of my jobs were about data. I analysed carbon footprint data at Publicis Sapient, cleaned up map data at GraphGeo, and spent most of 2025 building the backend for an AI interview platform.
Outside work I’m building Vele, a calendar you can talk to, and Juggle, a gateway that juggles my AI keys. I also write about how AI models work, and buy more Hot Wheels than I have space for.
- Glasgow
- Centre for Data Science and AI, University of Glasgow
- MSc Data Science at Glasgow, BE in AI and Data Science in Karnataka
- Prompt Corpus, a free course on prompting
- Vele, a calendar that takes instructions, and Juggle, a gateway for AI keys
- Hot Wheels
Things I’ve built.
Run down the list. Whichever one you land on plays.
Power BI dashboards you can download.
Some use public data, some use datasets I just found fun. Download any of them and open them in Power BI. They’re MIT licensed, so do what you like.
Better prompts, not bigger models.
When an AI model gives a bad answer, most people’s first move is to try a bigger one. That costs more money and more energy, and a lot of the time it barely helps.
Usually the problem is the question. If you tell the model who the answer is for, what format you want and what to avoid, an ordinary model will beat a bigger one that got a vague prompt. That’s why I think getting better at prompting is the more sustainable way to use AI.
Prompt Corpus
A free course I wrote on prompt engineering. It starts with how language models work, then builds the techniques on top of that.
It’s free to read. You make an account so it can keep track of where you are.
Read Prompt Corpus- 00Foundations12
- 01Beginner8
- 02Intermediate7
- 03Advanced9
- 04Model Playbooks6
- 05For Students7
- 06At Work5
- 07Cost of a Prompt6
98 sources, about 6h of reading.
A calendar that takes instructions.
You write it a sentence — “lunch with mom, 1 to 3 on Saturday” — and it pencils the event in. Anything with other people in it, or a clash, or a repeat, stays in pencil until you say yes.
The AI model reads the sentence. The times come from a separate scheduling engine that knows my working hours, buffers and lunch.
What the scheduling engine takes into account:
- Working hours
- Buffers
- Lunch
- Focus time
- Notice
- Daily limits
- Time zones
Your diary, where Vele writes in pencil and only you turn it into ink.
A gateway that juggles your API keys.
Free-tier AI keys are rate limited, so you end up holding a handful of them and swapping between them by hand. Juggle does the swapping for you. Your keys go in its settings, and your app calls one OpenAI-compatible endpoint.
It uses a different key for each request. When one gets throttled, the request moves on to the next key, so your app still gets an answer.
It works with Gemini for now. Other providers can be added as adapters.
Where I’ve worked and studied.
Now
University of Glasgow
From Oct 2026 / Glasgow
Administrator, Centre for Data Science and AI
- Admin work for the university’s data science and AI centre.
Before
University of Glasgow
Mar 2026 to Oct 2026 / Glasgow
Reach Out Ambassador, James McCune Smith Learning Hub
- Helped students and staff around the learning hub. I moved here from the IT ambassador job in March.
University of Glasgow
Jan 2026 to Mar 2026 / Glasgow
Reach Out IT Student Ambassador
- Helped students and staff with IT problems.
- Showed students and staff what the university’s IT services offer.
Inclusive Homework Club
Oct 2025 to Sep 2026 / Glasgow
Volunteer Tutor
- Spent a year helping students with English and maths.
- Got tutor of the week twice.
Technobeez Solutions
Jan 2025 to Sep 2025 / Remote
Lead Developer, EduDiagno
- Led the backend and AI work on EduDiagno, a platform that runs job interviews with AI.
- Built the live speech-to-text and sentiment analysis, which ran across more than 1,000 interviews.
- Set the backend up to handle 50 or more video interviews at the same time.
Publicis Sapient
Sep 2024 to Dec 2024 / Bangalore
Data Analyst Intern, Sustainability
- Worked with large sustainability and operations datasets for client projects.
- Built KPI dashboards for teams across the business.
Publicis Sapient
Nov 2023 to Dec 2023 / Bangalore
AI Engineering Intern
- Went through carbon footprint data for clients and found ways to cut their emissions by 10 to 15 percent.
- Built data pipelines that handled over 100,000 metrics a day.
- Made dashboards that cut manual reporting time by more than half.
GraphGeo Analytics
Jun 2023 to Aug 2023 / Bangalore
Data Analyst Intern
- Cleaned up and checked more than 10,000 map data records.
- Wrote scripts that cut tagging mistakes by about 40 percent.
Education
University of Glasgow
2025 to 2026 / Glasgow
MSc Data Science
- Machine learning, statistics and working with big datasets, finished off with a dissertation.
SDM Institute of Technology
2021 to 2025 / Karnataka
BE Artificial Intelligence and Data Science
- Machine learning, deep learning, data analysis and software engineering, mostly learned by building projects.
Skills
Languages
- Python
- SQL
- TypeScript
- JavaScript
- C
- Java
Analysis
- Exploratory analysis
- Hypothesis testing
- Time series
- Feature engineering
Machine learning
- Scikit-learn
- PyTorch
- TensorFlow
- Classification
- Clustering
- Prompting
Data engineering
- Data pipelines
- PostgreSQL
- MySQL
- Data modelling
- FastAPI
Dashboards
- Power BI
- DAX
- Power Query
- Tableau
Other tools
- Git
- React
- REST APIs
- Linux
Outside work.
What I spend my evenings on.
Hot Wheels
I collect Hot Wheels, the little 1:64 ones. The shelf filled up a while ago. I’m still buying them.
Pull the car back and let go.
Gaming
Genshin Impact and Red Dead Redemption 2 at the moment. Very ready for GTA VI.
Vibe coding
I build a lot of small apps by prompting AI. Most get thrown away. A few turn into something.
Cooking
I cook a lot. Traditional recipes, strange fusion experiments, and some baking when I’m in the mood.
Reading
Right now it’s the Shiva Trilogy by Amish Tripathi. Usually fantasy, sci-fi or a bit of philosophy.
Tutoring
I tutored English and maths at the Inclusive Homework Club for a year, and got tutor of the week twice.
Rather just ask?
Ask the chatbot anything about me. It answers from what’s on this page.
Hi. Ask me anything about Shreyas: where he’s worked, what he’s built, or what he does for fun.
Get in touch.
Want to talk about something, or know a Hot Wheels car I should hunt down? Send me a message.
