Vidyut Salaria
Background
The foundation is academic: an MSc with a dissertation that required building original research infrastructure from nothing, identifying a gap, designing the methodology, developing the platform, and producing results that held up under scrutiny. That work demonstrated something more useful than subject knowledge: the capacity to take an untested idea through a full project lifecycle and deliver a defensible outcome.
The projects that followed were each built on a defined business rationale. A problem with no adequate solution, a user segment underserved by existing tools, or a revenue model that had not been applied to a particular context. None were speculative. Each one was scoped, executed, and progressed as far as the current stage allows. The documentation here reflects that process in full: from the initial case for building, through to what the data showed once real users were involved.
The work
Six self-initiated builds. Six domains entered specifically because they were unfamiliar. The distance from each field was the analytical advantage.
As AI becomes more common as a workplace collaborator, one question had not been tested. Can AI trigger emotional contagion the way human teammates do. This matters most in high-stakes environments like healthcare and finance, where emotional states directly affect performance and decision quality.
Survey-based studies existed measuring self-reported emotional responses. Human-to-human teaming experiments had established methodology and documented findings. Both were well-documented and addressed different questions entirely. Neither examined human-AI collaboration in real time.
A web platform was built from scratch to run the experiment, designed to recruit participants internationally and capture performance data live across three separate conditions. Sixty-six participants completed the study. Those who knew they were working with AI scored approximately 87% higher on performance measures than those who did not. Statistical testing confirmed the results were reliable to a very high degree of confidence, with AI awareness accounting for roughly a third of the entire variation in performance outcomes. Awarded Distinction. Examiners described it as publication-level work.
The initial approach considered was a survey study, collecting questionnaire responses and analysing self-reported data. Ruled out because surveys measure what people report feeling, not what they actually do under pressure. Human-to-human teaming was also considered but set aside. That area had existing research behind it, though a full experimental study in real-time collaborative conditions had not been done. The harder question was chosen because it was the one that had not been answered.
Organisations introducing AI into team environments need to design those systems to support performance without becoming a source of emotional stress. The research extends social facilitation theory into human-AI teams and surfaces the limits of what AI can replicate from human interaction. Future work should test more adaptive AI systems across longer timeframes and more culturally diverse samples.
Approach
Precision framing
Every problem here was studied before it was touched. The goal was not familiarity but understanding where existing solutions had fallen short. The domains vary. The approach does not. Each assumption is tested against evidence, and only what holds is built upon. That is what separates the projects here from solutions that address the surface without resolving what caused it.
Identifying the underlying problem
Every project here started as something else on the surface. The expiry feature existed on every shop till. The forecast line existed on every stock application. Neither had been treated as an unresolved design problem. The work is always one layer deeper than what is immediately visible.
Full cycle product thinking
Product thinking here runs end to end, covering opportunity identification, solution positioning, route to customer, and revenue logic. StockPulse was preceded by a 15-section PRD covering user roles, business logic, and workflow before any code was written. A feature is not complete when the user is satisfied. It is complete when the business case closes.
Technical fluency, non-technical clarity
Complex technical concepts are absorbed quickly and translated accurately for non-technical audiences. The translation is not a simplification. It is a different precision: finding the explanation that is both correct and immediately usable by someone without the technical background.
Consistent delivery, variable conditions
Working mode is determined by the problem. Independent when it requires deep focus and full ownership. Collaborative when it requires range and shared expertise. The output standard does not change between the two.
Working principles
The positions that have stayed consistent across every project.
01 — AI that augments human judgement is more valuable than AI that replaces it. That was always the more interesting question.
02 — Convenience determines adoption. A feature that adds friction does not exist in any meaningful sense.
03 — Identifying what the analysis is missing matters more than perfecting what it already contains.
04 — Every claim requires a traceable source. Decisions built on unverifiable information carry risk that is never accounted for.
05 — A solution that requires technical literacy to operate has not solved the original problem. It has moved it.
06 — Tools built for everyone are optimised for no one. Specificity is where utility begins.
Areas of interest
Areas of ongoing interest outside the projects:
How businesses actually operate and scale. The decisions that determine whether expansion holds or breaks, how companies enter new markets, and what operational choices separate sustainable growth from collapse.
How geopolitical events reshape industries. Trade policy, regulatory shifts, and power realignments determine which technologies get built and which get blocked.
Robotics and physical AI. Where software intelligence meets the physical world and what that transition looks like at an industry level.
Deep tech as a category. The companies building at the layer below the application, where the barriers are highest and the timelines are longest.
The Toolkit
Everything I've picked up.
Tools and technologies used across research, development, and strategy.
Python
SQL
C
C++
HTML
CSS
JavaScript
Flutter
Android Studio
Git
GitHub
Docker
NumPy
PandasMatplotlib
TensorFlow
PyTorch
OpenCV
JupyterSPSS
sklearn
Plotly
Kali Linux
Wireshark
Metasploit
BurpSuiteAircrack-ng
John the RipperNmapHydra
HashcatNetcat
HackTheBox
Cisco Packet TracerBeEF
Agile/ScrumA/B TestingUAT TestingWireframingP&L BasicsPricing Strategy
Jira
AsanaProjectLibre
TrelloMonday.com
Miro
Notion
Google Analytics
HubSpotPower BITableauPower Query
Coursera
Udemy
edXLinkedIn LearningMIT OpenCourseWare
freeCodeCamp
HackerRank
Flask
Next.js
FastAPI
Node.js
Tailwind CSSPostgreSQL
MongoDB
MySQLMS Access
VercelHeroku
Render
Railway
Google CloudAWS
Cloudflare
Claude
GeminiGroq
PerplexityMidjourneyOpenAI APIAnthropic API
Hugging FaceGitHub Copilot
DuckDuckGo Search API
OllamaStable DiffusionWhisper
MarketlineMintel
Statista
Google Scholar
ScopusElsevierUK Gov Data
TradingView
Google WorkspaceSlackMicrosoft Teams
Zoom
Google Forms
BlenderCapCutCertifications
Committed to learning formally, not just building.
Formal credentials across product, cloud, data, and strategy.
Featured
Certified Scrum Product Owner (CSPO)
Scrum Alliance
McKinsey Forward Program
McKinsey & Company
EA Product Management Job Simulation
Forage · Electronic Arts
Siemens Mobility PM Job Simulation
Forage · Siemens
AWS Academy Cloud Foundations
Amazon Web Services
Also completed
Google Project Management Certificate
Coursera · Google
Oct 2021
Architecting with Google Compute Engine Specialisation
Coursera · Google Cloud
Oct 2020
Six Sigma Yellow Belt
Coursera
Lean Six Sigma White Belt
Management and Strategy Institute
Oct 2021
What's next
Open to the next challenge.
Currently based in the UK. Open to relocation. Open to the right role.