Applied AI · Analytics · Software
I build systems that help people make better decisions.
Audio-to-emotion models, AI scouting platforms, engagement analytics for universities — the common thread is turning messy data into something a person can act on.
Liam Shaw
02 / Work
Research that shipped. Products in active use. Click any row to open the full case study.
Multimodal Music Emotion Analysis
Graduation research project. A segment-level emotion recognition system that jointly models audio spectrograms and lyrics to track how a song's emotional character evolves over time — deployed as a full Flask web application with recommendations.
Generative AI Football Scouting Platform
Full-stack NLP platform for professional football analytics. Plain-language queries via NL2SQL, chain-of-thought AI scouting reports, ML player profiling with PCA and Random Forest, and four agentic capabilities — all over a 370 MB football database.
RoBERTa + 26 AMR Semantic Features
A hybrid NLP classifier combining RoBERTa transformer embeddings with 26 hand-engineered Abstract Meaning Representation features. Outperforms pure-transformer baselines by ~18 percentage points on FakeNewsNet (23,196 articles).
Client Work · FLAME Centre for Digital Learning
Deployed analytics dashboard for FLAME University's Centre for Digital Learning. Replaced a manual Excel-based workflow — CSV export, manual scoring, spreadsheet flagging — with a seconds-to-insight web application tracking pre-orientation student engagement.
Additional work
ResNet-18 + DistilRoBERTa · Gated Fusion
4-quadrant emotion classifier using gated late-fusion of ResNet-18 audio encoding with DistilRoBERTa lyric encoding. The sigmoid gate dynamically weights each modality per sample — Q2 (Angry) F1 = 0.91.
01 / About
I'm a final-year Computer Science student at FLAME University, Pune (Minor: Business Analytics, Full Academic Scholarship). My work sits at the intersection of applied ML, NLP, and software engineering — with a consistent focus on outcomes over elegance.
My graduation project, MoodTrack, is a multimodal music emotion system that models audio and lyrics jointly at sub-second resolution. The most interesting result wasn't the CCC scores — it was that the learned fusion gates independently recovered a known music-psychology finding purely from gradient descent, without any explicit supervision. That's the kind of result that earns trust.
At ACEplus, I designed the analytics event framework for a product reaching 5,000+ schools, ran cohort analyses that directly shifted product strategy, and built a data-cleaning pipeline from scratch. At FLAME's CDL, I replaced an hours-long manual Excel workflow with a deployed client-side dashboard in weeks — by choosing the right constraint early.
I'm drawn to roles where both technical rigour and practical judgment matter: AI engineering, product analytics, CRM/ops tech, and full-stack software.
Every AI project I build ends with a running application. MoodTrack is a Flask web app. ScoutAI is a full-stack platform. Results tables alone don't tell you whether a system is useful.
CS + Business Analytics background means I always ask: what decision does this data change? The 2× cohort finding at ACEplus mattered because it shifted product strategy — not because it was statistically interesting.
The client-side Moodle dashboard shipped in weeks because building local-first removed IT approval and privacy objections in one move. The right constraint often makes the project possible.
I move across NLP research, backend engineering, CRM analytics, and client work. The common thread: understand the problem clearly, pick the right tool, execute cleanly.
Liam Shaw
FLAME University, Pune
CS + Business Analytics · 2026
03 / Experience
ACEplus · Derek O'Brien & Associates · Kolkata · Hybrid
Mandrake Tech · Pune
Willwali · Remote · Kuala Lumpur
FLAME Sports · FLAME University, Pune
04 / Skills
ML & Deep Learning
PyTorch, scikit-learn, CNN-BiLSTM, ResNet-18, XGBoost, PCA, KMeans, Random Forest
NLP & Transformers
RoBERTa, DistilBERT, HuggingFace, NL2SQL, RAG, AMR Parsing, Semantic Search
Languages
Python, Java, JavaScript, SQL, C++, HTML/CSS
Frameworks & Infrastructure
FastAPI, Flask, Spring Boot, React, Docker, Node.js, SQLAlchemy
Analytics & CRM
CleverTap, Cohort Analysis, Funnel Analysis, pandas, NumPy, Matplotlib, FuzzyWuzzy, Excel
Tools & Platforms
Git, Jira, SQLite, PostgreSQL, Bubble.io, Jupyter, HuggingFace Hub
05 / Education
FLAME University, Pune
La Martiniere for Boys, Kolkata
Certifications
CS50x: Introduction to Computer Science
HarvardX · edX · 2021
Building AI Literacy
LinkedIn Learning · Mar 2025
Responsive Web Design
freeCodeCamp · 2021
06 / Contact
I'm actively looking for internships and full-time roles in AI engineering, analytics, data products, CRM/ops tech, and software. If you're working on something interesting — or you just want to talk about a project — reach out.
© 2026 Liam Shaw · liamshaw.in
Next.js · Tailwind · Framer Motion