= |
> Akuna Capital Quant Trading Insight '26
> Campus Marketing Lead, Adobe Express
> EMEA Cloud Student Programme @Google
> Winter Tech Fellowship @Palantir
> Code in Place @Stanford
> Mathematics for ML @Imperial
> Google DeepMind: AI Research Foundations
$ █
Second year Computer Science undergrad. Perpetually curious about how things are engineered and how they're governed, which is how I've found myself prototyping distributed systems in one fortnight and dissecting trading strategy in the next. Stanford Code in Place, Wharton Global Youth Program, a summer at Brown, Akuna's quant trading insight programme, Harvard Model United Nations (HMUN) among several outstanding delegation awards back at school, rooms I sought out because sitting still was never really on the cards. I've been an ambassador at Adobe Express, Canva, ThAT, and a fair few others along the way, convening summits and conferences, leading and speaking in equal measure. Beyond the stage, my leadership takes a quieter form: mentoring first year students in reaching the goals I once chased without a guide. And still, somehow, on the lookout for the next room to walk into.
First Class Honours
Activities and societies
Linear Algebra 98%, Multivariate Calculus 94.75%
Activities and societies
Major Projects
Fellowships & Programmes
Published Papers
Led Adobe Express's on-campus marketing strategy as part of Adobe's Student Ambassador Program, owning outreach, content, and society partnerships end to end.
Selected for Akuna Capital's competitive Quantitative Trading Challenge, gaining hands-on exposure to trading strategy and market dynamics within a small, selective cohort.
An immersive day with HSBC: candid conversations with recruiters and graduate hires, followed by mock interviews and bespoke feedback to sharpen interview technique.
Ranked 11th, exploring AI and software architecture with the Manchester-based team.
(During winter break from school) Selected to intern with Palantir, building an AI solution on their AIP platform under the mentorship of Palantir engineers.
(During summer break from school) Selected for Stanford's Code in Place, an intensive introduction to programming taught by Stanford faculty and section leaders.
Selected for Wharton's Global High School Investment Competition, developing and pitching a simulated investment portfolio as part of a cohort of student teams.
Helping host seminars and conferences across Europe.
Upcoming: π¬π§ AE Global Summit on Open Problems for AI 2026 in November.
Selected for Cohort 001 of DisCo, The Discovery Community, built around scientific discovery, reasoning, and problem-solving. Excited to be in the room with researchers and thinkers pushing on the same questions.
Selected into Canva's second UK cohort of Gen C Student Ambassadors, recognised for campus leadership in creative technology.
Part of a dedicated team of students volunteering countless hours between lectures and exams, curating the "Rethinking Tomorrow" theme, sourcing speakers, and orchestrating every logistical detail from scratch, all while juggling the pressures of our own projects and assignments.
Selected from a competitive pool of applicants as a Peer Mentor for first year Computer Science students, entrusted with regular check ins, coursework guidance, and a steady, approachable presence, helping them achieve their goals and make the most of their time at uni.
Jun 2026
View credentialLinkedIn Learning
Jun 2026
View credentialJPMorganChase
Jun 2026
View credentialGoogle DeepMind
Jul 2026
View credentialSkyscanner
Jul 2026
View credentialDeloitte
Jul 2026
View credentialLive data from @disha-umeshaiah
A short collection of things I've built.
Vector search engine in Python. HNSW implemented from the paper, recall within 1β2pp of FAISS's production implementation.
MiniVDB is a vector search engine written from scratch in Python, implementing Hierarchical Navigable Small World (HNSW) graphs directly from the original paper rather than wrapping an existing library. The goal was to understand approximate nearest neighbour search at the level of the underlying graph structure, not just call an API.
Benchmarked against FAISS, a production grade vector search library, MiniVDB's from scratch HNSW implementation reaches recall within 1 to 2 percentage points of FAISS's own numbers, using NumPy for the underlying vector operations.
Distributed cache in Go. Consistent hashing, SWIM gossip, custom binary protocol β nodes detect and route around failures in ~560ms, no coordinator.
ShardCache is a distributed in-memory cache written in Go, built around consistent hashing to distribute keys across nodes and a SWIM style gossip protocol for cluster membership, so nodes can detect and route around failures without relying on a central coordinator.
Failed nodes are detected and traffic rerouted in roughly 560 milliseconds, using a custom binary protocol for inter-node communication rather than an off the shelf serialisation format.
This site, built from scratch with scroll-triggered animations, a dark and light mode toggle, glassmorphism elements, and a neon visual identity. No frameworks.
This site, the one you are looking at, is built entirely from scratch with no frameworks. Scroll triggered animations reveal each section as you move down the page, and a dark and light mode toggle switches the entire colour system, including the glassmorphism panels used throughout.
The neon visual identity, from the glowing text to the terminal styled hero, was built with hand written CSS rather than a component library, alongside a working command palette and live GitHub activity integration pulling real contribution data.
A focus timer with a four-screen state machine, CSS parallax animations, and an ambient sound engine synthesised from a sawtooth oscillator using the Web Audio API. Includes a real-time terminal logger. Zero dependencies.
AeroFocus reframes a standard Pomodoro style timer as a four screen state machine, moving between setup, focus, break and summary states with CSS parallax animations that shift depth as you move through a session.
Rather than a static alarm, the ambient background sound is synthesised live using the Web Audio API from a sawtooth oscillator, so no audio files are loaded and the tone is shaped entirely in code. A real time terminal logger runs alongside the timer, printing session events as they happen.
A Chrome extension that steps in at checkout with a brief prompt before a purchase completes, aimed at reducing impulse spending in the moment it happens.
NeonFlow Pro is a Chrome extension that intervenes at the exact moment a purchase is about to happen. Rather than a finance app you check after the fact, it surfaces a brief prompt right at checkout, aimed at giving a second of pause before the click that completes the order.
Built directly on the Chrome Extension API, it hooks into checkout pages without requiring a separate app or dashboard, keeping the intervention as close to the decision point as possible.
Browser game engineered during Google's EMEA Cloud Student Programme using AI-integrated development tools under an accelerated timeframe.
Built during Google's EMEA Cloud Student Programme, this browser based game was engineered inside an accelerated, deadline driven timeframe using Google AI Studio's AI integrated development tools rather than a conventional local setup.
Working directly against API keys within the programme's environment meant designing for rapid iteration from the outset, since there was no room for a slow build and test cycle within the time given.
If youβre looking to collaborate on research, have an interesting challenge in mind, or simply want to connect, Iβd be glad to hear from you.
dishaumeshaiah@icloud.com