Different problems. One approach.
My work has taken me from public-service delivery to road safety, education, elections, policy research and software. The domains change. The work is similar: understand the problem, make the data usable, find what matters, and turn it into something people can use. Pick a domain.
Speed Safety Score: are the limits even right?
Road safety is usually framed around drivers who exceed the limit. We asked a different question: does the posted limit make sense for the road itself? Our three-person team scored road segments across Maharashtra and Thailand from 0 to 100, then ranked where intervention would matter most.
Try the idea: does the limit fit the road?
Illustrative only. A simplified Safe System speed logic to show the concept, not the competition model.
Observe
Start from ADB-provided GPS speed data: how fast vehicles actually travel on each segment.
Enrich
Layer in population, infrastructure, settlement and street imagery to understand who is exposed.
Score
A 0–100 score per segment: Safe System alignment, speed-limit credibility gap and vulnerable-road-user risk.
Prioritise
Rank roads for intervention under limited budgets, using exposure, likelihood and severity.
A road network, scored segment by segment
Illustrative network, not real data. Hover a segment. Scores use the same simplified logic as the explorer above.
From a fifth of the network to all of it
GPS speeds cover only part of the road network. A machine-learning model (XGBoost) extends the score to the rest.
Hidden-danger roads
Drivers obey the limit, yet the limit itself is unsafe. They are invisible if you only count speeding.
Recommendations in plain English
Each road gets an action a transport authority can read and act on.
What if citizens could report a problem in their own words?
GovLens AP is a prototype exploring how multimodal AI could make grievance reporting easier, and turn unstructured complaints into records that can be routed and analysed.
A citizen sends whatever they have
No forms to decode. Voice notes are transcribed and translated automatically.
The AI turns it into a structured record
The right desk sees it, and the citizen sees progress
Officers get an action log and analytics. Scheme details are verified through web search.
A personal finance product built around cash flow.
Don't just show people where their money went. Help them understand what to do next.
Built around your salary cycle
Months rarely match paydays, so the app tracks from payday to payday.
Bill scanning that stays on your phone
On-device OCR reads a bill; the photo is never uploaded to an AI service.
A financial health score you can inspect
A 0–100 score with a transparent breakdown, not a black box.
Actions ranked by financial impact
Recurring bills on their real due dates, spending forecasts and a pace chart.
Experiments and studies.
Smaller projects where I wanted to try an idea: multi-agent AI, signal processing and classical machine learning.
mst-abs
Extracting structure from noisy atmospheric radar signals: adaptive Bayesian shrinkage applied to MST radar data.
Car Price Prediction
A comparative study of decision trees, random forests, neural networks and polynomial regression for used-car price prediction, with tuned hyperparameters.
Seven roles across government, a university, research and technology.
Each one started with a messy question and ended with something people could use: a model, a dashboard, a measurement system.
Consultant, Power Sector Analytics
Analytics and evaluation work in the power sector.
Project details are confidentialBusiness Analyst
The data behind a 10,000+ candidate scholarship exam, and a university's budget decisions.
- Managed operational data for Pathway 2026, a President's Office initiative to find and support meritorious students from marginalised communities: about 150 exam centres, 10,000+ registered and 8,000+ appearing candidates.
- Built revenue and scholarship scenario models from historical admissions data and stress-tested them against prior-year enrolment. Leadership adopted one scenario to set budget allocations.
- Produced governance-level performance reports and diagnostics to find where programmes were falling short.
Data Scientist
Dashboards and spatial analysis for district-level governance programmes.
- Consolidated multi-department administrative data into Power BI dashboards and structured reports for state leadership.
- Mapped forest cover and land use in QGIS, computing land-composition shares at constituency and district level.
- Integrated and validated datasets from several departments to support programme outcome review.
- Deployed and tested KoboToolbox for field data collection, including a non-containerised installation.
Data Consultant
Performance analysis across SDG, tourism and investment programmes.
- Combined quantitative and qualitative evidence to support programme evaluation and strategic planning.
Project Lead Analyst
Made a state grievance system measurable.
- Led monitoring and analytical reporting for Spandana, the state's public grievance redressal system.
- Co-developed SOPs and KPIs with senior officials so that programme monitoring became routine, not ad hoc.
- Designed executive dashboards to track service delivery, surface bottlenecks and support administrative decisions.
Senior Data Analyst
Analytics tooling for policy research and field reporting.
- Built automated data-collection pipelines to support policy research and programme assessment.
- Built dynamic Plotly Dash dashboards on Heroku for field reporting.
Data Analyst
Electoral and demographic data across three election cycles.
- Analysed demographic and electoral datasets to support policy research and field evaluation.
- Designed analytical dashboards and carried out extensive field data collection in southern Rajasthan.
Where evidence meets human outcomes.
Policy writing as a research volunteer with Our Gender Lens.
Understanding the Female Happiness Paradox in India
Policy Brief No. 1
Gender, Child Development, and the Gap Between School and Work in India
Policy brief
Data & analytics
From messy administrative data to something you can analyse.
Statistics & machine learning
Methods for asking whether something worked, and what comes next.
Visualisation & geospatial
Dashboards and maps that a decision-maker can read in a minute.
Software & AI
Turning an analysis into a tool other people can use.
Research & policy
Design and measurement for public programmes.
Data platforms
Where the data lives and how it is collected in the field.
Mechanical engineer by training. Data person by practice.
I started in mechanical engineering, moved into data analytics, and gradually found myself working where technology, public policy and real-world systems meet.
I'm drawn to problems where the data is messy, the question isn't fully defined, and the answer needs to be more than a chart. Sometimes that's a model. Sometimes a dashboard, a policy brief or a product.
I'm less attached to a particular tool than to understanding the problem and building something useful.
Have a problem worth solving?
I'm interested in analytical, policy, infrastructure and technology problems where better evidence can lead to better decisions.