Abhivir Singh
22, from Faridabad. Studying AI & ML at Imperial College London (MEng Computing, final year). Before that, IIT Delhi. Before that, a lot of competitive exams.
I like building things end-to-end. I like hard problems and I like shipping.
Graduated in 2026. Back in India, building.
Highlights
See all highlights →
Imperial College London →
2022–2026MEng Computing (AI & ML). Final year.
academic
IIT Delhi, B.Tech
2021Admitted for B.Tech through JEE Advanced, top 0.1% nationally. Transferred to Imperial CS after one year to specialise in AI.
academic
Cambridge ENGAA: 9.0/9.0
2021Perfect score. Unconditional offer from Trinity College, Cambridge University.
academic
Imperial Cricket Captain →
2023–2025Led the team in BUCS. Scored a 100, earned the baggy blue cap. Batting avg: 81.
leadership
YC Top 25: Dinner with YC Partners
Apr 2026Top 25 of 25,000 builders selected for an exclusive dinner with YC partners Jared Friedman, Jon Xu, Ankit Gupta, and Arnav Sahu (Peak XV). Writeup coming.
event
Winner: Wintermute Crypto CTF ↗
2026Self-taught Zero Knowledge Proofs and elliptic curve cryptography to compete and win.
competitionFeatured Work
See all work →AI Data Science Platform →
2024–2025Solo-built platform with nine AI agents that coordinate across the full data science pipeline. Humans in the loop at every decision point. Live with data scientists at Fetch.ai.
AI for Fetal Heart Ultrasound ↗
Oct 2025 – Jun 2026MEng thesis. MobileUNet for real-time fetal heart segmentation, AutoFHR for heart rate estimation, L-FUSION for probe navigation. Targeting portable devices for low-resource settings.
Pintos OS & WACC Compiler →
2023–2024Operating system and compiler, both from scratch. Best in cohort of 230 at Imperial. 95% mark.
Emotional Text-to-Speech →
Winter 2024/25Fine-tuned open-source TTS to express emotion. Compiled an emotional audio dataset with 2,000+ HuggingFace downloads.
GPT from Scratch
Summer 2024Transformer language model built from scratch, following Karpathy's NanoGPT. Multi-head self-attention, BPE tokenizer, training and inference.
MEGA: Multilingual LLM Evaluation ↗
Autumn 2024Research paper evaluating LLMs across 70+ languages. Identified data contamination issues in benchmarks and evaluated compression parity as a fairer metric.
Right Now
What I'm up to →Graduated from Imperial. Building AI agents and the harnesses around them. In India, working out where to do the next serious thing.
Get in Touch
Want to work together, have a question, or just want to say hello?