Available for opportunities · London, UK
I build AIproducts thatrun inproduction.
Voice agents, predictive models, full-stack apps.
AI / software engineer with an MSc in Software Engineering (Distinction) from the University of West London, and a year of commercial React and Django work for a US healthcare client. Lately: a real-time AI phone receptionist and a deep-learning predictive-maintenance system for UAV fleets.
Based in
London, United Kingdom
01 / Conversational AI
AI Receptionist
A production-ready AI receptionist that answers business calls 24/7, captures booking details through natural conversation, and emails structured leads to the owner. Built on Twilio voice + OpenAI — available to integrate into your business.
Result: Live demo on this site, 100+ automated tests, now being demonstrated to prospective UK pub customers.
02 / Deep Learning
UAV Predictive Maintenance Digital Twin
LSTM, Transformer, and 1D-CNN models forecasting Remaining Useful Life on NASA C-MAPSS turbofans and a synthetic UAV fleet — with an interactive 3D dashboard and explainable-AI diagnostics.
Result: Transformer reached 14.12 RMSE on held-out NASA C-MAPSS engines and 97.4% fault-classification accuracy on the UAV fleet.
Based in
London, United Kingdom
What I do
AI Applications & Voice Agents
Real-time voice agents and LLM features built on OpenAI, Deepgram, ElevenLabs and Twilio — with structured outputs validated against strict JSON Schemas.
Backend & APIs
TypeScript and Python services with Node.js, Express, FastAPI and Django REST Framework, backed by PostgreSQL and Prisma, tested with Vitest and shipped through CI/CD.
Machine Learning
PyTorch models for prediction and fault diagnosis, evaluated on held-out data to avoid leakage and explained with Integrated Gradients and attention maps.
Full-Stack Web Apps
React and Next.js front ends styled with Tailwind CSS, from interactive dashboards to the product pages that put AI in front of people.
More work
All projects →Airfoil Selector for Wind Turbines
A tool that ranks airfoil designs for small vertical-axis wind turbines by annual energy production at a specific site — pairing real wind statistics with NeuralFoil surrogate aerodynamics and generating explainable reports on why one design outperforms another.
Guitar Tablature Generation
LSTM and feed-forward neural network that predicts optimal guitar tablatures for any input melody, modelling musical intention and fretting difficulty.
Music Genre Classification
CNN model classifying audio into 10 genres using MFCC features from the GTZAN dataset, with custom preprocessing.
Result: 79% accuracy across 10 genres on GTZAN.
Kind words
“Swarnim is an exceptional professional who brings genuine passion and dedication to everything he undertakes. His commitment to excellence and ability to inspire and uplift his team members makes him an invaluable asset to any project.”
“After facing significant academic challenges, Swarnim's patient tutoring and dedication transformed my educational journey. Under his mentorship, I not only passed but gained confidence in both Nepali and Chemistry. His approach made complex subjects accessible.”
Have a projectin mind?
Open to junior and graduate software, backend and AI engineering roles — and to businesses that want an AI receptionist.