My work splits across two fields that job adverts usually keep apart. Each project below is described twice — once as behavioural research, once as impact analysis. Same work, different reading. Use the switch at the top to pick one.
2025 – 2026MSc dissertation · LSE
Sustainability on the Move
Pro-environmental practice and horizon security under chronic liminality · supervised by Dr Atrina Oraee
Behavioural read
Sequential mixed-methods study of international postgraduates who had moved home more than three times in three years: 65 eligible survey responses screened from 181, then 12 interviews analysed with reflexive thematic analysis. Built on the habit discontinuity hypothesis, social practice theory and Installation Theory. There was no uniform erosion — practices moved in different directions, and how easy something was to do predicted whether it happened far better than how much the person cared did.
Impact read
Explains why recycling, energy and food habits collapse in short-let and shared housing, even among people who care. What decides whether a practice survives a move is whether the new place makes it workable, and whether the resident can see far enough ahead to bother setting it up. That puts the problem with landlords, universities and housing providers, not with tenant education campaigns.
What the data showed
Across seven practices there was no consistent direction of change. Sustainable transport rose from 64.6% to 98.5% and was the only result robust to every sensitivity check — though destination infrastructure is confounded with the move itself, which the text says plainly. Energy saving fell from 72.3% to 55.4% and food-waste reduction from 76.9% to 63.1%; neither survived both robustness checks, and I report them as such. Almost a third of people who had recycled at home had stopped.
Perceived ease correlated with frequency across all five separately measured practices (ρ = .46 to .67, every one at p < .001). Environmental self-identity correlated with only two of them. Identity travels with the person; the arrangement decides whether the practice actually happens. The most common reason given for stopping was that the facilities were not there (52.4%), and just under a quarter said it was not worth it while the living situation was temporary.
The contribution
A dependency–horizon model: a 2×2 separating how far a practice depends on its physical setting from how certain a person is about that setting lasting. Horizon security splits into duration certainty (how long do I have) and succession certainty (do I know what comes next), and the two behave differently.
Feb 2023 – presentUBC Visual Cognition Lab
The Correlation Project
Advisor since Jan 2026; previously Project Leader, Secondary Project Leader and Co-Pilot · PI Dr Ronald Rensink
Behavioural read
Psychophysics of how precisely and accurately people perceive correlation in scatterplots, using just-noticeable-difference discrimination and direct estimation, across 300+ participants and 18+ conditions. My own COGS 402 study tested five opacity levels and five mark shapes. Correlation judgements survived every one of them except crosses.
Impact read
Impact and ESG reporting assumes the reader takes the right relationship out of a chart. This work measures how far that holds when the design changes — mostly it does, which is the useful finding for anyone choosing chart styles. Three years at the lab, two of them running the project, and then a handover.
What the shape and opacity study found
Precision and accuracy tracked each other across all five opacity levels, as Fechner's assumption predicts: the discrimination and estimation offsets were statistically indistinguishable in every sub-condition, and the Weber slope varied by only 0.03 across them. Four of the five mark shapes behaved the same way. Crosses did not — there the two offsets came apart (p = 0.025), and the shape condition showed an effect the opacity condition did not (p = 0.013 against 0.27). So correlation perception is largely invariant to how you draw the marks, with one specific exception worth knowing about if you are designing a chart. Sixty-plus participants run over two months; 55 produced usable datasets across the two conditions, and 39 survived the exclusion criteria for the estimation task.
How it ran
Experiments run on a locally hosted Node.js application, extending the lab's existing JavaScript codebase. Analysis in Excel using Spearman and Pearson correlation with sensitivity checks. I recruited, onboarded and trained nine research assistants, mentored two students through independent projects, and ran three concurrent experimental conditions. I now advise the project rather than lead it.
Feb 2026LSE Specialist Projects: Sustainability
Funding strategy for EWCPA
Student consultant, in an LSE team, to a Kenyan NGO working with imprisoned mothers and their children
Behavioural read
Worked out which parts of the programme depended on particular people and local knowledge, and which could be written down and repeated somewhere else. Then reframed how the organisation talks about itself — from listing activities to showing outcomes, which is what funders actually respond to.
Impact read
Benchmarked EWCPA against four peer organisations to find the niche it could defend, mapped a funding portfolio across institutional consortia, entry-level grants, diaspora sponsorship and faith-based donors, and built a three-tier impact scorecard so the organisation could report outcomes at the standard institutional funders expect.
What we recommended
Stop competing with larger organisations on economic empowerment, and position instead as a specialist sub-grantee on the mother–child relationship — the one thing no peer in the sector centres. The diligence also turned up a live legal risk: Kenya's Public Benefit Organisations Act retired the old NGO category in 2024, which put eligibility for international funding in question until the organisation's status was confirmed. We sequenced the work so the nearest grant deadline came first and the evidence base needed for larger institutional funders was built behind it.
2023 – 2025Gaurav Impex, Mumbai
Procurement and operations
Silicone and rubber raw materials · distributor for DJ Silicone, plus EPDM and neoprene grades
Behavioural read
Rebuilt how the business talks to its customers: a Zoho CRM setup with a custom Trials module tracking material samples through to decision, and a branded email template system sent from the CRM. I still maintain both. Two years of watching what buyers respond to is a good corrective to lab findings.
Impact read
Two years inside a materials supply chain: sourcing, supplier terms, pricing and stock across silicone, EPDM and neoprene grades. Sustainability analysis of manufacturing inputs reads differently once you have done the procurement yourself and know where the costs sit.
What it changed
Diagnosed sales logistics delays as misaligned incentives, then designed feedback loops and OKR-style goal setting across a 12-person team, cutting order turnaround from one working day to two hours. Restructured supplier negotiations across a £1M portfolio using anchoring and loss-aversion framing, for a 5% cost reduction — roughly £40,000 a year. Rebalanced resource allocation across three functions using demand analysis on sales data, moving customer acquisition from a run-rate of 50–75 to 100+ over three months.
Earlier: Tutor, The C.O.D.E. Initiative · Content Writer, TeenBook India
Full write-ups of the dissertation and the correlation study are available on request.
