Pradhyumna Shashi Bhushan — AI/ML Engineer & Technical Lead

Issue 01 / Vol. 2026

00 — ContentsIn this issue

In this issue

01PrologueWho, and why
02Selected WorkFive features
03ExperienceRoles & impact
04Research & SchoolingPapers, degrees
05SkillsTools & technologies
06CorrespondenceWrite to me

01 — PrologueWho, and why

A note from the engineer

Hi, I'm Pradhyumna — an AI and Software Engineer with over two years of experience in LLMs, VLMs, AI/ML engineering, computer vision and cloud architecture. I've led enterprise AI/ML delivery on Azure, worked on diverse projects across India and handled multiple international ones — including AI solutions for ICAR, IIMR and Roquette, ONDC integrations, and EMR platforms for aesthetic clinics in Dubai. I enjoy building practical AI that solves real-world problems across industries.

Currently

Reading for an M.Sc. in Autonomous Systems at the University of Stuttgart. Open to AI/ML engineering roles and a few select freelance builds.

02 — Selected Work(05) features

Selected work,
2025 — 26

Three production platforms built for Quant Innovations FZE in Dubai — shipped to the web, the App Store and Google Play — and two projects from research and my own studio.

Feature N° 01 / 05Quant Innovations FZE, Dubai2025 — 26

Snaptank

Every guest, every photo — found by face and delivered to WhatsApp before the party ends.

01A face-recognition engine that matches a single guest selfie against thousands of event photos in seconds.
02Instant delivery to WhatsApp and email, with QR registration, watermarking, sponsor branding and live event analytics.
03A web dashboard plus iOS and Android apps, subscriptions, an admin console and a partner programme.

Role: Lead engineer — web, mobile, ML

Visit snaptank.ai↗

Fig. 01Guests at a sunset yacht party — the kind of event Snaptank delivers photos for

Snaptank campaign artwork: a camera scattering instant prints — Fig. 01Guests at a sunset yacht party — the kind of event Snaptank delivers photos for

Feature N° 02 / 05Quant Innovations FZE, Dubai2026

Asttona

The quiet precision of a modern clinic — an EMR and CRM for aesthetic and dental practices.

01One platform for every clinic: patient records, digital consent, before/after imaging, bookings, billing, inventory and a patient portal.
02A private AI assistant for clinic staff — patient data stays inside the clinic, and staff approve every action it takes.
03Built to UAE healthcare standards, with encryption, multi-factor sign-in and a tamper-evident audit trail.

Role: Full-stack & AI engineer

Visit asttona.com↗

Fig. 02The Asttona clinic dashboard on a laptop and phone in a calm clinic interior

A clinician reviewing a patient record on a tablet — Fig. 02The Asttona clinic dashboard on a laptop and phone in a calm clinic interior

Feature N° 03 / 05Quant Innovations FZE, Dubai2026

LetsDecide

Ending the “I don't know, you decide” loop — friends swipe together and the app finds the unanimous match.

01Real-time swipe rooms for movies, restaurants, recipes and plans — with tie-breaks, leaderboards and match celebrations.
02“Rhea”, a voice-enabled planning assistant that suggests what to do next based on the weather and your calendar.
03Bill splitting from a photo of the receipt, eleven languages including right-to-left Arabic, live on iOS and Android.

Role: Sole engineer — product to stores

Visit letsdecide.app↗

Fig. 03A dim cinema with movie posters — LetsDecide's movie-night category

Feature N° 04 / 05Research project2025

Swarm SAR

Search-and-rescue drone swarms that learn to coordinate — and to save battery — on their own.

01Multi-agent reinforcement learning with Proximal Policy Optimization for swarm coordination in obstacle-laden terrain.
02Reward shaping and state-space optimisation for battery-aware search strategies.
03Over 90% mission success across diverse simulated scenarios.

Role: Reinforcement-learning engineer

Simulation — move your cursor through the swarm

sim / ppo-swarm-v2

agents 16

targets found 00t = 0

Fig. 04Search-and-rescue drone swarms that learn to coordinate

Feature N° 05 / 05Independent product2025

OpusAI

An interview coach that listens, then raises the bar — adaptive voice interviews and resume-aware job matching.

01Adaptive voice interviews that calibrate question difficulty from real-time feedback — a 50% lift in session completion.
02NLP resume parsing and job matching that surfaced ~15 relevant roles per candidate in their first week.

Role: Founder & engineer

Demo on request

Live interview — round 03difficulty ▲ adaptive

“Walk me through a system you designed end to end.”

Fig. 05An interview coach that listens, then raises the bar

03 — Experience2024 — now

Work experience

From an internship automating LLM pipelines to leading enterprise AI delivery — two and a half years, three desks, one habit: ship it to production.

- 01Led architecture and end-to-end delivery of enterprise AI/ML solutions on Microsoft Azure — machine learning, data engineering, cloud infrastructure and production deployment.
- 02Set technical strategy, architecture standards, security controls and deployment practice across the AI/ML and cloud stack.
- 01Built a multilingual agentic LLM with Retrieval-Augmented Generation and BiLSTM forecasting that gives farmers real-time maize prices and connects buyers across 12+ regions.
- 02Integrated ONDC and eNAM to link buyers and sellers across the agricultural supply chain — increasing market reach by 30%.
- 01Built and released Python LLM pipelines on Clarifai that automated workflows and lifted operational efficiency by more than 20%.
- 02Designed a Chrome extension that scraped structured data from 10+ websites, cutting manual collection time by 70%.

04 — Research & SchoolingPapers, degrees

On the record

PublicationIEEE SSITCON 2025

NeuroSphere: Optimizing Mental Wellness Through Reinforcement Learning

A peer-reviewed paper on applying reinforcement learning to personalised mental-wellness support.

DOI 10.1109/SSITCON66133.2025.11342145

Read the paper↗

Schooling

2026 — Present
University of Stuttgart
M.Sc. Computer Science — Autonomous Systems
Stuttgart, Germany
2022 — 2026
Dayananda Sagar College of Engineering
B.E. Information Science & Engineering
CGPA 9.39 / 10

05 — SkillsTools & technologies

Languages & ML

Python
TypeScript
PyTorch
TensorFlow
LangChain
AutoGen
Java
C

Product engineering

Next.js
React
FastAPI
WebSockets
Capacitor
Stable-Baselines3
OpenCV
Hugging Face

Data & cloud

Azure
AWS
Docker
PostgreSQL
MongoDB
Redis
Firebase
NeonDB
Vercel
Prometheus

06 — CorrespondenceOpen to roles & select freelance

Let's build
something intelligent.

Write
to me

Direct line

pradhyumna.sb@gmail.com