Skip to content
Mantrana logo Mantrana

Internships

Internships

Mantrana internships are research projects with a mentor, a weekly rhythm and real deliverables. Pick a track, check the skills each project needs, and tell us in the form what you would bring.

Social Welfare Analytics A 16-week project on one public health or welfare programme: read the scheme documents with AI, build a systems model, design and run a survey, and test what would help over five years. AI in Health Care Clinical prediction models and hospital AI projects with SSSIHMS: explainable machine learning, language models, speech and scheduling for day-to-day hospital work.

Apply now How to fill in the form

For everyone

Skills every intern needs

You do not need all of these on day one, but you should be comfortable starting with them.

Python

Jupyter, pandas and scikit-learn for data work; reading and fixing other people's code.

Git

Code, prompts and data versions kept in a repository so the work can be reproduced.

Careful reading

Reading long programme documents and clinical protocols closely, and noting the page for every claim.

Clear writing

Weekly progress notes, a project charter, and a final brief written for people who are not specialists.

Using AI tools responsibly

Language models speed up the work, but every AI output is checked by a person and logged.

Ethics and privacy

Consent, anonymising data at source, and keeping sensitive records on approved systems (DPDP Act 2023).

Track 1 · Social Welfare Analytics

Skill groups and the projects that use them

Every social welfare project uses the first three groups. The others depend on the project, so look for the groups you are strongest in.

  1. 1 Mapping the Hypertension Care Cascade
  2. 2 Diabetes Care in Practice
  3. 3 From Policy to Plate
  4. 4 Midday Meal Delivery Under the Lens
  5. 5 Strengthening Palliative Care Delivery
  6. 6 Containing Foot-and-Mouth Disease and Lumpy Skin Disease in Cattle

Systems modelling

All projects

Causal loop diagrams, stock-and-flow models, calibration to past data, and scenario testing.

Vensim or Python (PySD, SciPy), NetworkX, SALib, Optuna

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6

Document AI and text extraction

All projects

Parsing scheme PDFs, extracting causes, targets and timelines with language models, and scoring the results against hand-coded pages.

Docling, PyMuPDF, OCR (Tesseract, PaddleOCR), LLMs with JSON output, precision, recall, F1

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6

Survey research and fieldwork

All projects

Turning model links into survey items, translation and back-translation, piloting, reliability checks and interviews with frontline workers and families.

KoboToolbox or ODK, CFIR and RE-AIM measures, Cronbach's alpha, factor analysis

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6

Machine learning and statistics

4 projects

Risk and dropout prediction with explanations, time-to-event models, trend forecasting and anomaly detection.

XGBoost or LightGBM with SHAP, Cox models, Prophet or Chronos, Isolation Forest

  • 1
  • 2
  • 3
  • 6

Computer vision

4 projects

Image models as research aids, such as retinal screening, child measurement, plate waste and skin lesions in cattle.

Fine-tuned ViT, ConvNeXt, DINOv2 or YOLO; image segmentation

  • 2
  • 3
  • 4
  • 6

Speech and Indian languages

3 projects

Transcribing voice interviews and analysing themes and sentiment in Indian languages. Speaking the local language of the study area helps.

Whisper, AI4Bharat Indic ASR, IndicTrans2, MuRIL or IndicBERT

  • 1
  • 5
  • 6

Optimisation

1 project

Linear programming for menus that meet nutrition norms, and searching for the best bundle of interventions.

Linear programming, Optuna or BoTorch

  • 4

Epidemic and spatial modelling

1 project

Outbreak and vaccination dynamics, Bayesian calibration and maps of where disease spreads.

SEIR models, PyMC, geopandas, Moran's I

  • 6

Which project suits you

1 · Chronic disease

Mapping the Hypertension Care Cascade

Students interested in chronic disease, health systems and prediction models.

Systems modellingDocument AI and text extractionSurvey research and fieldworkMachine learning and statisticsSpeech and Indian languages

Project details

2 · Chronic disease

Diabetes Care in Practice

Students who like modelling disease progression; image work if the scheme includes eye screening.

Systems modellingDocument AI and text extractionSurvey research and fieldworkMachine learning and statisticsComputer vision

Project details

3 · Nutrition

From Policy to Plate

Students interested in child nutrition, community services and data quality.

Systems modellingDocument AI and text extractionSurvey research and fieldworkMachine learning and statisticsComputer vision

Fieldwork involves children: ethics approval, parental consent and child assent.

Project details

4 · Nutrition

Midday Meal Delivery Under the Lens

Students who enjoy optimisation, school systems and analysing audit reports.

Systems modellingDocument AI and text extractionSurvey research and fieldworkComputer visionOptimisation

Surveys with children need ethics approval, parental consent and child assent.

Project details

5 · Palliative care

Strengthening Palliative Care Delivery

Students drawn to qualitative research, caregiving and language AI.

Systems modellingDocument AI and text extractionSurvey research and fieldworkSpeech and Indian languages

Interviews with caregivers use trauma-sensitive protocols.

Project details

6 · Animal health

Containing Foot-and-Mouth Disease and Lumpy Skin Disease in Cattle

Students interested in epidemiology, animal health, maps and rural fieldwork.

Systems modellingDocument AI and text extractionSurvey research and fieldworkMachine learning and statisticsComputer visionSpeech and Indian languagesEpidemic and spatial modelling

Project details

16 weeks

How the 16 weeks run

The social welfare projects follow the same ten stages, from scope to recommendations.

  1. Week 1

    Scope and ethics

    Ten ICMR videos with a viewing log for each, a one-page project charter, a document inventory and the ethics checklist.

  2. Weeks 2–3

    Read the documents

    Parse the scheme PDFs, hand-code a gold set of pages, and extract causes, targets and timelines with AI.

  3. Weeks 4–5

    Draft the model

    A causal loop diagram, checked with officers and frontline workers.

  4. Weeks 6–8

    Build the survey

    Items from the model, expert review, translation, cognitive interviews and a pilot.

  5. Weeks 9–12

    Fieldwork and analysis

    Surveys and interviews, cleaning, theme analysis, and a link-by-link gap analysis.

  6. Weeks 13–16

    Model

    Calibrate and validate the model, test interventions, and write the policy brief.

Every week

Monday
Plan the week: three goals and the deliverable due on Friday.
Wednesday
30-minute check-in with your mentor.
Friday
Submit the deliverable and a short note: done, blocked, next.

Checkpoints at weeks 5, 8, 11 and 14 confirm the model, ethics, data and validation before moving on.

What you will hand in

  • Viewing logs and a one-page project charter
  • A hand-coded gold set and an extraction pipeline with its accuracy scores
  • A variable dictionary, causal loop diagram and stock-flow model
  • Survey instruments, codebook and sampling plan
  • A gap analysis of the programme on paper versus on the ground
  • A calibrated model with validation, sensitivity and scenario results
  • A policy brief with recommendations and indicators to monitor
  • A Git repository that lets someone else reproduce the work

Track 2 · AI in Health Care

AI in Health Care

Work with clinicians at SSSIHMS on the applications in development and the hospital AI projects on the Research page.

Clinical prediction

Explainable machine learning on clinical data, such as the brain abscess, mitral stenosis and lumbar disc projects.

Python, scikit-learn, XGBoost, SHAP, careful handling of patient data

Language models in the hospital

Triaging radiology reports, sorting patient email, and answering statistics questions in plain language by writing the database query.

LLM prompting and testing, SQL, retrieval

Speech and documentation

Turning outpatient conversations into notes, and voice-driven tools for the radiology information system.

Whisper or other speech-to-text, summarisation

Scheduling and operations

CT and MRI scheduling, staff rosters, queues and room planning.

Optimisation, constraint solving, simulation

See the clinical projects

Apply

Filling in the application form

What each question on the form is looking for.

Before you apply

  • Read the project you want on the Social Welfare Analytics or Research page
  • Know which skill groups you bring and which you want to learn
  • Resume uploaded to Google Drive and shared with "Anyone with the link"
  • GitHub or portfolio link that opens without signing in

Open the form

  1. 01

    Full name, email and mobile

    The details we use to contact you. Use an email you check daily.

  2. 02

    Institute, campus and department

    Where you study now. Department is optional.

  3. 03

    Current year of study

    I to IV UG, PG or other, so we can match the work to where you are.

  4. 04

    Area of interest

    Choose Social Welfare Analytics or AI in Health Care. If you have a particular project in mind, choose Other and name it.

  5. 05

    Why do you want to intern at Mantrana?

    The most important answer. Name one or two projects, the skill groups you are strongest in, and something you have built or studied that shows it. A few specific lines beat a long general one.

  6. 06

    Resume link

    A Google Drive link to a PDF. Set sharing to "Anyone with the link can view", or we cannot open it.

  7. 07

    LinkedIn, GitHub or portfolio

    GitHub is best if you have code to show: a notebook, a model, a scraper or a survey analysis. Otherwise LinkedIn or a portfolio.