Sanya Upadhyay

Data Scientist

3years in AI/ML

6/10Accessibility & Semantics

5years in tech

2years leading teams

Overview

Data-focused engineer with experience delivering Python-based data products and machine-learning workflows using SQL and Azure Databricks. Built and operationalized a Random Forest classifier, implemented LLM-based extraction pipelines with privacy-aware data handling, and created analytical dashboards for research and operational decision-making. Also has prior MATLAB/Simulink development experience for collision-risk algorithms.

Phone

Timeline

Data Engineer · Middle

Civic Health Innovation Labs, University of Liverpool · Full-Time

Apr 2025 to Present 1 Year 7 Months Liverpool In office

Developed and tested Python and PySpark data products on Azure Databricks, orchestrated via Azure Data Factory and supported by automated tests and Git-based collaboration. Deployed a Random Forest model into a production Python workflow and improved model quality by analyzing cross-validation metrics and prediction errors. Built a vacancy-analysis pipeline combining deterministic parsing, locally hosted LLM extraction, and classification to select reliable extraction methods. Prepared linked clinical and survey datasets for longitudinal research, supported responsible data handling, and contributed to an ecosystem spanning a React Native patient app, a Next.js clinician portal, a .NET API, and cloud storage.

Python

SQL

pySpark

Scikit-learn

Azure

Azure Data Factory

Delta Lake

NumPy

LLM

Ollama

Pytest

Git

GitHub

Docker

React Native

Next.js

.NET

KTP Associate – Data Engineer · Middle

WMG, University of Warwick / Gordon Ellis · Full-Time

Mar 2024 to Apr 2025 1 Year 1 Month In office

Delivered an Innovate UK-funded digital transformation project by coordinating academic, business, and engineering stakeholders and managing technical milestones and risks. Built Python ETL pipelines and designed a normalized SQL database to integrate energy sensor data with different sampling intervals. Applied regression-based baselines and anomaly detection to identify inefficiencies and out-of-hours consumption patterns. Created Dash/Plotly dashboards to translate analysis into operational decisions and improve data availability through failure diagnosis.

Python

SQL

Plotly

Anomaly Detection

University of Liverpool

Master's Degree · Data Science & Artificial Intelligence

2022–2023 Liverpool, England

Software Developer · Middle

Tata Consultancy Services (TCS) – Ford Germany · Full-Time

Jan 2021 to Sep 2022 1 Year 8 Months In office

Developed and validated collision-risk algorithms using MATLAB and Simulink, working with radar/LiDAR inputs and time-to-collision logic. Implemented numerical modeling and sensor-data filtering, then performed scenario-based testing and debugging to improve algorithm reliability. Automated recurring engineering tasks using a MATLAB App Designer tool, reducing repetitive effort for the development workflow.

MATLAB

Simulink

App Designer

Pune University

Bachelor's Degree · Electronics & Telecommunication Engineering

2016–2020 Pune, Maharashtra

Technical skills

Languages

MATLAB

Python

SQL

JavaScript

AI/ML

PyTorch

TensorFlow

LLM

NumPy

Scikit-learn

Jupyter Notebook

Ollama

Reinforcement Learning

Pandas

Spark

DevOps

Azure

GitHub

Docker

Git

Other

Expo

Simulink

App Designer

Plotly

Next.js

React Native

Delta Lake

Pytest

.NET

pySpark

Azure Data Factory

React.js

Databricks

Anomaly Detection

Machine Learning

AI Agents

Frontend

Agentic Workflows

Post-training

ETL/ELT

Synthetic Data

Middle Frontend Developer Confidence: Medium UI Engineer

Middle-level UI engineer focused on accessible, multilingual front-end interfaces with an emphasis on audio-first interactions. The strongest proven skill is robust client-side integration of text-to-speech and localization, demonstrated by app.js features such as speechRun-based race handling, AbortController-backed server TTS cancellation, speechChunks splitting, and translateLooseText translation logic. There is little to no public evidence of test coverage, build/tooling automation, large-scale architecture or production-grade backend services beyond a simple server TTS endpoint.

UI Component Architecture

3/10

How interface parts are built

Modular vanilla-JS functions and a small reusable floating audio control show deliberate separation of concerns, but there is no component framework, design-system abstraction, or clear component API surface that would indicate higher-level component architecture work.

Evidence

app.js: getVoicesReady / chooseVoice / speechChunks / speak - modular function boundaries

index.html: floating-audio controls wired to speak/pauseSpeak/stopSpeak (reusable UI block)

app.js: document.querySelectorAll('.show').forEach(b=>b.onclick=...) - small UI behavior composability

Responsive & Cross-browser

4/10

Works on all screens and browsers

Practical cross-browser and localization concerns are handled explicitly (viewport meta, RTL toggling, speech API feature detection), but there is limited use of advanced responsive strategies like container queries or fluid typography.

Evidence

app.js: applyLang sets document.documentElement.dir and toggles .rtl class for RTL locales

index.html: meta viewport for responsive layout and language select with aria-label

app.js: feature detection for 'speechSynthesis' and voiceschanged handling

Performance Optimization

5/10

Speed of the interface

Meaningful performance-minded code is present for the specific problem space: chunking long text to avoid TTS failures, timeout-based voice loading, and revoking object URLs to prevent leaks. There is no repository-level instrumentation, bundle splitting, or measured before/after analysis.

Evidence

app.js: speechChunks(text,max=180) - splits long text into safe TTS chunks

app.js: getVoicesReady(timeout=1800) - voiceschanged event plus timeout fallback

app.js: stopSpeak revokes cloudAudioUrl (URL.revokeObjectURL) to avoid resource leaks

Accessibility & Semantics

6/10

Usable for everyone

Good accessibility attention for an HTML/JS prototype: ARIA role=status live updates, labeled controls, form validation/reportValidity, visible-language and RTL support. Some advanced a11y patterns (keyboard focus management for custom widgets) are minimal but present patterns are correct and pragmatic.

Evidence

app.js: speechStatus creates element with role='status' and updates live messages

index.html: language <select> has aria-label and floating-audio container has aria-label

activities.html: form.demo-form uses required fields and f.reportValidity() with a visible success UI

State Management & Data Flow

5/10

Managing data in the app

Concrete async state discipline is implemented: cancellation with AbortController for server TTS, speechRun counter to avoid race conditions between overlapping speak requests, and clean stop/pause logic; this shows practical experience with client-side async flows though not a full state-machine or formal client cache strategy.

Evidence

app.js: speak/cloudAbort + AbortController to cancel in-flight /api/tts fetch requests

app.js: speechRun counter used to ignore stale async results and stop in-progress browser speech

app.js: translatedPageText() builds a bounded text payload (slice(0,5000)) for server requests

UX & Visual Polish

5/10

Look and feel quality

UX details are well considered for the product: visible status messages, fallback notifications when voices are missing, playback-rate control, form success states and localized 'show more' toggles. Visual polish is prototype-level and not backed by animation standards or complex micro-interaction system.

Evidence

app.js: speechStatus messages and localized fallback notices when no installed voice is found

index.html: floating controls include playback speed select and pause/resume behavior

index.html & pages: 'Show More' toggles wired to simple expand/collapse behavior in app.js

Expertise

Frontend AI Integration· Middle

HTML & CSS· Middle

Industries

Health Care· Middle

Technologies

Frontend

Next.js· since 2025

Azure· since 2025

React.js

Recommendations

Build accessible, multilingual public-facing web interfaces that require robust client-side audio and TTS workflows.
Implement prototypes integrating cloud neural TTS (Azure) with client-side cancellation, fallback and localized UX.
Improve developer impact by adding automated tests (unit + integration), CI, and a minimal build/tooling pipeline.
Work on audio-first web experiences and accessible content-savvy components for community or nonprofit sites.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

Hackathon-Team-LetsTalkTdoday

Job preferences

Work permits: China, India, United Kingdom
Citizenship: India
Salary: from $70k per year
Employment type: Full-Time
Locations: Manchester, London

Open to AI agents: they read this profile and start a dialog through Alionalion.io/@sanya-upadhyay/mcp