# MyChild Engine — Full Site Content

> Concatenated markdown content of every page on www.mychildapp.in, for AI agents and LLMs. See /llms.txt for the index. All validation numbers are from synthetic data — NOT clinical validation. MyChild Engine is not a diagnostic tool.

---

# MyChild Engine — Open-Source Developmental Screening

> An open-source screening engine that tracks developmental milestones across 8 domains, from birth to 36 months. Every question is weighted by clinical evidence, every result is traceable to the exact rule that produced it, and nothing ever leaves your device. NOT a diagnostic tool.

- Install: `npm install mychild-engine` — Apache-2.0 (code) / CC BY-SA 4.0 (data)
- GitHub: https://github.com/hsongra11/mychild-app
- Canonical HTML page: https://www.mychildapp.in/

## The problem

The free checklists from the CDC don't score anything. The validated tools like ASQ-3 and M-CHAT-R/F do, but they're copyrighted and expensive. There's nothing in between. MyChild Engine fills that gap — it tracks milestones over time, weights each observation by clinical evidence, and tells you exactly why it flagged something. Not a diagnosis. Just enough signal to know when to talk to your pediatrician.

## Under the hood

- **131 weighted questions** — Each question carries a weight: Low, Medium, High, or Red-Flag. Low-weight items need corroboration before they matter. Red-flags skip the line.
- **8 independent domains** — Gross Motor, Fine Motor, Receptive Language, Expressive Language, Social-Emotional, Cognitive, Self-Help, Vision/Hearing. Each scored on its own.
- **10 age-appropriate bands** — Birth through 36 months, plus universal red flags that apply at any age.
- **False alarm protection** — The engine won't flag "high concern" from a single observation. It requires at least 2 independent data points before escalating.
- **Preterm correction built in** — For children born before 37 weeks, the engine automatically adjusts expectations until 24 months.
- **Show your work** — Every result comes with a plain-English explanation of exactly what drove it.
- **Rule simulator** — Run synthetic child timelines against different threshold configurations and see exactly which alerts shift.

## Validation study (synthetic, not clinical)

Verified against 71,939 synthetic developmental profiles generated by an autoresearch system (inspired by Karpathy's autoresearch):

| Metric | Value | Notes |
|---|---|---|
| Cohen's Kappa | 0.906 | Combined set; above 0.8 is "almost perfect" in clinical research |
| Sensitivity | 91.4% | Share of real concerns the engine catches |
| Specificity | 98.8% | Share of typically-developing children correctly cleared |
| Profiles tested | 71,939 | Synthetic profiles — not real children |

Two phases: a hand-verified set (1,747 profiles, kappa 1.000 by construction) and an adversarial stress test (70,192 profiles designed to break the engine — kappa 0.852, specificity 98.0%). Ground-truth labels came from an AI clinical evaluator, not licensed clinicians. **This is not clinical validation.**

Full study: https://www.mychildapp.in/study.md · Raw data: https://www.mychildapp.in/study-data.md

## How it compares

| Tool | Sensitivity | Specificity | Validation |
|---|---|---|---|
| ASQ-3 | 82–97% | 83–93% | Clinical (~18,000 children) |
| M-CHAT-R/F | 85–95% | 93–99% | Clinical |
| PEDS | 74–79% | 70–80% | Clinical (~1,500 children) |
| MyChild Engine (combined) | 91.4% | 98.8% | **Synthetic only** |

Every other tool has been validated with real children; MyChild Engine hasn't. The comparison is directional, not a claim of equivalence.

## India / RBSK alignment

India has 26 million babies a year. RBSK (Rashtriya Bal Swasthya Karyakram), India's national child health screening program, is validated at 97% sensitivity and 96.4% specificity against ASQ-3. MyChild Engine's question bank is mapped domain-by-domain against RBSK screening items — alignment for integration planning, not endorsement or validation. Details: https://www.mychildapp.in/rbsk.md

## What's next

1. Adaptive questioning — stop asking once the engine has enough signal
2. NLP probe analysis — understand free-text parent answers
3. Longitudinal pattern recognition — distinguish late bloomers from persistent delays, catch regression

## Recognition

Built by Harsh Songra and Aafreen Ansari — Forbes 30 Under 30 (India & Asia), TEDx speakers, 500 Startups portfolio, WEF Agenda contributor. Google filmed a documentary about the original MyChild App (https://www.youtube.com/watch?v=5h4wcz2zbWA). Story: https://www.mychildapp.in/about.md

## Quick start

```bash
npm install mychild-engine
```

```js
import { evaluate, computeChildAge, getDueQuestions } from 'mychild-engine';

// Get age-appropriate questions for a 7-month-old
const questions = getDueQuestions({ dob: new Date('2025-09-01') }, []);
// Returns 12 questions across Motor, Language, Cognitive, Social
```

Docs: https://www.mychildapp.in/docs.md · npm: https://www.npmjs.com/package/mychild-engine

## FAQ

**Is MyChild Engine clinically validated?**
No. The engine has been verified against 71,939 synthetic developmental profiles (Cohen's kappa 0.906, 91.4% sensitivity, 98.8% specificity), but synthetic profiles are not real children. A prospective clinical study with real patients is the next milestone, and we are actively looking for clinical partners.

**Is MyChild Engine a diagnostic tool?**
No. It tracks developmental milestones and surfaces patterns so caregivers know when to talk to a pediatrician. It cannot and does not diagnose any medical condition, developmental disorder, or disability.

**How much does MyChild Engine cost?**
Nothing. It is fully open source: the code is Apache-2.0 and the question bank and study data are CC BY-SA 4.0. Install it with `npm install mychild-engine`.

**What ages does MyChild Engine cover?**
Birth to 36 months, across 10 age bands, plus universal red flags that apply at any age. For babies born before 37 weeks, the engine automatically uses corrected age until 24 months.

**How is my child's data protected?**
The engine runs entirely locally on your device. No child data is transmitted to any server. This website uses standard analytics for page views only — see https://www.mychildapp.in/privacy.md for details.

**How does the scoring actually work?**
The engine asks 131 evidence-weighted questions across 8 developmental domains. Each question carries a weight class (Low, Medium, High, or Red-Flag) that controls how fast it escalates. Evidence sufficiency gates require at least 2 independent observations before flagging high concern, and every result includes a plain-English explanation of exactly what drove it.

**How is this different from CDC checklists or ASQ-3?**
CDC milestone checklists are free but have no scoring logic. Validated instruments like ASQ-3 and M-CHAT-R/F are scored but copyrighted and expensive. MyChild Engine sits in the gap: free, open source, and scored — though unlike ASQ-3 and M-CHAT-R/F it has not yet been clinically validated.

**What should I do if the engine flags a concern?**
Book an appointment with your pediatrician and bring the engine's explanation trace with you — it shows exactly which observations drove the flag. If your child loses skills they previously had, seek medical advice promptly. For anything urgent, contact your local emergency services.

## Where the questions come from

1. Zubler JM, et al. "Evidence-Informed Milestones for Developmental Surveillance Tools." *Pediatrics*. 2022;149(3). doi:10.1542/peds.2021-052138
2. Lipkin PH, Macias MM. "Promoting Optimal Development…" *Pediatrics*. 2020;145(1). doi:10.1542/peds.2019-3449
3. CDC "Learn the Signs. Act Early." milestone checklists — cdc.gov/act-early/milestones
4. RBSK (Rashtriya Bal Swasthya Karyakram) — rbsk.mohfw.gov.in
5. Synthea synthetic data validation — doi:10.1186/s12911-019-0793-0
6. "Synthetic data can aid the analysis of clinical outcomes." *PNAS*. 2024. doi:10.1073/pnas.2414310121

## Important

**This is not a diagnostic tool.** It tracks developmental milestones and surfaces patterns. It cannot and does not diagnose any medical condition, developmental disorder, or disability. If something concerns you about your child's development, talk to your pediatrician. Thresholds are hypothesis-level and have not been validated through clinical trials. The engine runs entirely locally — no child data leaves your machine.

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# About MyChild Engine

> How a 9-year misdiagnosis turned into an open-source developmental screening engine used across 100+ countries. Built by Harsh Songra and Aafreen Ansari.

Canonical HTML page: https://www.mychildapp.in/about

## The story

In 2015, Aafreen Ansari and Harsh Songra raised $100K from 500 Startups and angel investors to build MyChild App. Parents in over 100 countries used it to screen their kids for developmental delays. Sheryl Sandberg wrote about it; Forbes put both founders on 30 Under 30 lists. The startup eventually wound down, but the core problem didn't go anywhere.

Ten years later, the engine is open source — the rules, the question bank, the scoring logic. All of it.

## Where this started

Harsh was diagnosed with dyspraxia at 11. It took 9 years for anyone to figure out what was going on — years between "something feels off" and "here's what it actually is." At 18 he built the first version of MyChild App; at 19, he and Aafreen dropped out of college to go full-time. The idea was never to replace doctors — it was to give parents a way to notice patterns early enough that the 9-year gap stops happening to other kids.

## Who we are

### Harsh Songra — Creator & Lead Engineer
Diagnosed with dyspraxia at 11 after 9 years of nobody catching it. Built the first MyChild App at 18. Raised $100K from 500 Startups and angels, spoke at TEDx twice, named Forbes 30 Under 30 in India and Asia.

- Forbes 30 Under 30 (India & Asia): https://www.forbesindia.com/article/30-under-30/30-under-30-harsh-songra-transforming-lives-of-children/42255/1
- TED talk: https://www.ted.com/talks/harsh_songra_why_do_we_fail_to_understand_people_with_disabilities
- WEF Agenda contributor: https://www.weforum.org/stories/authors/harsh-songra/
- LinkedIn: https://linkedin.com/in/harshsongra · Website: https://harshsongra.com

### Aafreen Ansari — Co-founder & Content / UX Lead
Dropped out of college at 19 to co-found MyChild App. As COO she shaped every word parents read and every screen they touched, raised seed funding from 500 Startups and Singapore Angel Network, spoke at TEDx, made Forbes 30 Under 30 Asia, and later co-founded We, Included, a mental health awareness platform.

- Forbes 30 Under 30 Asia 2017: https://www.forbes.com/pictures/58e7adec31358e58f07a36e9/songra--aafreen-ansari-20/
- LinkedIn: https://linkedin.com/in/aafreen-ansari-8a4265112

## Why open source, why now

The 131 questions, the evidence weights, the scoring logic, the age-band calibrations — that's years of real work and it shouldn't sit in a dead repo. Developmental screening shouldn't be locked behind expensive licenses. Open-sourcing the engine means any developer can put this in a clinic, a school, or a parent's phone. One developer, one weekend — that's the barrier now.

## Where we are now

The engine has been verified against 108 synthetic developmental scenarios (98.6% agreement across 294 hand-labeled observations, Cohen's kappa 0.971) and stress-tested with 71,939 synthetic profiles. It's a software verification study, not clinical validation. The question bank is mapped against India's RBSK national screening program. Next: adaptive questioning, NLP for free-text parent answers, and longitudinal regression detection.

## The people behind this

Investors: 500 Startups, Singapore Angel Network, Amit Gupta (InMobi), Arihant Patni, Anisha Mittal, Pallav Nadhani, Dr. Ritesh Malik, Lalit Mangal (Commonfloor).

Mentor: Samir Bangara (1977–2021) — "He taught me empathy and emotional intelligence. Always available. Always there to pick you up when you fall."

Always there: Luis & Fiona Miranda.

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# MyChild Engine — Autonomous Validation Study

> 71,939 synthetic profiles. Cohen's kappa 0.906, sensitivity 91.4%, specificity 98.8%. 14 engine improvements across 4 autonomous research runs. Every number is reproducible. This is a software verification study, NOT clinical validation — ground-truth labels came from an AI clinical evaluator, not licensed clinicians.

- Authors: Harsh Songra & Aafreen Ansari · Last updated: July 2026
- Clinical review: not yet — actively seeking clinical partners
- Canonical HTML page: https://www.mychildapp.in/study

## Headline metrics

| Metric | Value | Scope |
|---|---|---|
| Cohen's Kappa | 0.906 | Combined set |
| Sensitivity | 91.4% | Flags real concerns |
| Specificity | 98.8% | Doesn't cry wolf |
| Observations | 34,587 | Domain-profile pairs scored |

## Two-phase validation

**Phase 1 — Curated set (hand-verified):** 1,747 profiles, 12,991 observations, kappa 1.000, 100% sensitivity & specificity. Profiles promoted from the adversarial set after manual review confirmed unambiguous agreement. Kappa = 1.000 is expected by construction; this validates logic on clear-cut cases, not edge-case performance.

**Phase 2 — Adversarial stress test:** 70,192 profiles, 21,596 observations, kappa 0.852, sensitivity 86.5%, specificity 98.0%. Profiles designed to break the engine: regression patterns, sparse data, contradictory answers, preterm edge cases, single-domain flags. Of the 781 disagreements, most were under-flags — the engine is cautious.

## What "adversarial" means (18 categories)

Regression (10,578), Sparse data (9,072), Single domain (8,786), Mutations (6,500), Exhaustive age (5,616), Contradictory (5,208), Clear delay (5,073), Clear typical (4,964), Preterm (4,752), Borderline (2,897), Combinatorial (2,544), Real-world (1,914), and more.

**315 critical findings need human review** — adversarial profiles where the engine and the AI evaluator disagreed on clinically meaningful classifications. Logged and triaged, not hidden.

## Comparison with clinical tools

| Tool | Sensitivity | Specificity | Validation | Sample |
|---|---|---|---|---|
| ASQ-3 | 82–97% | 83–93% | Clinical | ~18,000 |
| M-CHAT-R/F | 85–95% | 93–99% | Clinical | — |
| PEDS | 74–79% | 70–80% | Clinical | ~1,500 |
| MyChild (adversarial) | 86.5% | 98.0% | Synthetic | 21,596 obs |
| MyChild (combined) | 91.4% | 98.8% | Synthetic | 34,587 obs |

ASQ-3, M-CHAT-R/F, and PEDS are clinically validated with real patient data. Ours are synthetic. You can't put them on equal footing.

## 14 improvements across 4 runs

- **Run 1:** isolated single-flag downgrade, cross-domain sparse correction, regression bypasses evidence gate, cross-domain regression protection
- **Run 2:** soft regression detection, multi-regression escalation, caregiver uncertainty detection, developmental sequence anomaly, asymmetric screening thresholds, + 5 more (RF scoring, label normalization, evidence gate tuning, low_concern handling, sparse precaution)
- **Run 3:** 8 refinement commits polishing edge cases
- **Run 4:** zero changes — 325 threshold experiments all returned "keep current parameters"

## Threshold optimization

The autoresearch system swept every tunable parameter (delta-not-yet interactions, delta-unsure values, infant and toddler grace periods, yellow/orange/red thresholds) — 325 experiments. Every single one came back "keep current parameters." Current values: T_yellow = 2, T_orange = 4, T_red = 6, grace 3w/12w. To improve from here we need to change algorithms, not knobs.

## RBSK alignment

RBSK screens 270 million children across India, independently validated at 97% sensitivity and 96.4% specificity against ASQ-3. Our question bank is mapped domain-by-domain against RBSK screening items. That doesn't make us RBSK-validated. See https://www.mychildapp.in/rbsk.md

## The original verification

Before the autoresearch system: 108 hand-labeled scenarios, 294 labeled observations, 98.6% agreement, Cohen's kappa 0.971, 4 threshold cases, 45 abstentions. Per-domain: Gross Motor (72 obs, 100%), Cognitive/Play (22, 100%), Self-Help (14, 100%), Vision/Hearing (23, 100%), Expressive Language (67, 98.5%), Social-Emotional (45, 97.8%), Receptive Language (34, 97.1%), Fine Motor (17, 94.1%). Every miss was a borderline over-flag — when the engine gets it wrong, it errs toward concern.

## The honest version

**What this shows:** kappa held at 0.852 under adversarial attack; 98.0% adversarial specificity; 14 algorithmic fixes shipped from autoresearch findings; 325 threshold experiments all said "keep"; question bank maps to CDC 2022 milestones and RBSK items; fully open source and reproducible.

**What this does NOT prove:** clinical validation with real patient data (needs IRB oversight and clinical partners); real-world sensitivity/specificity; cross-language/cross-culture performance; clinical-grade thresholds. 315 critical adversarial findings still need a clinician's review. A prospective clinical study is next.

## Reproduce it yourself

```bash
npm install mychild-engine
npx mychild-engine validate --profiles data/synthetic-profiles.json --format markdown
npx mychild-engine autoresearch --runs 4 --adversarial 70000
```

Research: https://github.com/hsongra11/mychild-app/tree/main/docs/research · npm: https://www.npmjs.com/package/mychild-engine

## Sources

1. Zubler JM, et al. *Pediatrics* 2022;149(3) — doi:10.1542/peds.2021-052138 (CDC 2022 revised milestones, 75th percentile standard)
2. Lipkin PH, Macias MM. *Pediatrics* 2020;145(1) — doi:10.1542/peds.2019-3449 (AAP surveillance & screening policy)
3. CDC "Learn the Signs. Act Early." — cdc.gov/act-early/milestones
4. RBSK — rbsk.mohfw.gov.in
5. Synthea validation study, *BMC Med Inform Decis Mak* 2019 — doi:10.1186/s12911-019-0793-0
6. "Synthetic data can aid the analysis of clinical outcomes," *PNAS* 2024 — doi:10.1073/pnas.2414310121
7. "Synthetic Validation of Pediatric Trust Instruments Using LLMs," MedRxiv 2025

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# MyChild Engine — Validation Study (Plain-Language Summary)

> Why MyChild is back, and what the validation numbers mean in plain language. 71,939 synthetic profiles, Cohen's kappa 0.906. Not clinical validation.

Canonical HTML page: https://www.mychildapp.in/study-simple · Full study: https://www.mychildapp.in/study.md

## Founder note

"I built the first version of this engine in 2014, when I was 16. Launched the app in 2016. It screens for developmental delays based on CDC milestones, WHO guidelines, and Red Flags indicators. I ran the startup for a couple years, then wound it down. Almost nobody is still working on this. The few who are aren't open-sourcing it. With what AI can do now, I decided to bring it back. I'm not raising money or restarting the company. I'm open-sourcing the engine so any platform that reaches parents can use it, improve it, and help push it toward clinical validation. The real goal is to get this validated and give it away." — Harsh Songra

## The numbers so far

- **Cohen's Kappa 0.906** — agreement between engine and evaluator beyond chance; above 0.8 is "almost perfect" in clinical research
- **Sensitivity 91.4%** — when there's actually a concern, the engine catches it 91.4% of the time
- **Specificity 98.8%** — when a child is developing normally, the engine correctly says so 98.8% of the time (1.2% false alarms)
- **71,939 profiles tested** — synthetic developmental profiles, not real children

## How we tested it

Phase 1: 1,747 hand-verified profiles (kappa 1.000 by construction — validates logic on clear-cut cases). Phase 2: 70,192 adversarial profiles designed to break the engine (kappa 0.852, specificity 98.0%; most of the 781 disagreements were cautious under-flags). Ground-truth labels came from an AI clinical evaluator, not licensed clinicians — this is not clinical validation.

## How it compares

ASQ-3 (82–97% sens / 83–93% spec), M-CHAT-R/F (85–95% / 93–99%), PEDS (74–79% / 70–80%) — all clinically validated with real children. MyChild Engine (91.4% / 98.8%) — synthetic data only. Directional comparison, not equivalence.

## Reproduce it

```bash
npm install mychild-engine
npx mychild-engine validate --profiles data/synthetic-profiles.json --format markdown
```

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# MyChild Engine — Raw Study Data

> Full validation dataset for the MyChild Engine verification study: 1,747 curated profiles, 70,192 adversarial stress-test profiles, manifest, and SHA-256 checksums. Open data, CC BY-SA 4.0. All profiles are synthetic — no real children.

Canonical HTML page: https://www.mychildapp.in/study-data (request download links by email there, or browse GitHub: https://github.com/hsongra11/mychild-app/tree/main/docs/research)

## What's in the dataset

- **Curated Profiles** (JSON, 1,747 profiles, ~8 MB) — hand-labeled developmental profiles plus promoted high-confidence adversarial profiles. Each simulates a child's milestone responses across 8 domains (1–24 months) with ground-truth labels and the engine's classification.
- **Adversarial Profiles** (JSON.GZ, 70,192 profiles, ~3.7 MB compressed / ~124 MB uncompressed) — profiles designed to break the engine: regression patterns, sparse milestone data, contradictory answers, preterm edge cases. Includes per-profile disagreement analysis and failure mode classification.
- **Data Manifest** (PDF, 17 KB) — dataset descriptions, file sizes, SHA-256 checksums, data provenance.
- **Complete Package** (ZIP, ~3.7 MB) — everything in one download.

## How the data was generated

1. **Synthetic profile generation** — an AI system generates developmental profiles simulating realistic milestone response patterns: a hypothetical child at a specific age (1–24 months) with answers across 131 evidence-weighted questions spanning 8 domains. Includes both on-track children and simulated delays of varying severity.
2. **Ground-truth labeling** — each profile is independently evaluated by an AI clinical evaluator assigning "no concern," "monitor," or "refer," using CDC 2022 / WHO guidelines through a different reasoning pathway than the engine. Not a gold standard — a licensed clinician would be — but consistent and reproducible.
3. **Engine classification** — the engine processes each profile through its deterministic, inspectable rule-based scoring.
4. **Comparison and analysis** — classifications compared to compute Cohen's kappa, sensitivity, specificity; disagreements analyzed into under-flags, over-flags, and threshold boundary cases.
5. **Adversarial stress testing** — increasingly difficult edge cases: regression after meeting milestones, sparse or contradictory data, preterm-adjusted ages, genuinely ambiguous boundaries. Most of the 781 disagreements come from here.

## Data format

```json
{
  "id": "syn-00042",
  "age_months": 14,
  "preterm_weeks": 0,
  "domains": {
    "gross_motor":     { "score": 0.82, "flags": [] },
    "fine_motor":      { "score": 0.65, "flags": ["pincer_grasp_absent"] },
    "communication":   { "score": 0.91, "flags": [] },
    "cognitive":       { "score": 0.78, "flags": [] },
    "social_emotional": { "score": 0.88, "flags": [] },
    "adaptive":        { "score": 0.73, "flags": [] },
    "sensory":         { "score": 0.95, "flags": [] },
    "feeding":         { "score": 0.80, "flags": [] }
  },
  "engine_result": "monitor",
  "ground_truth": "monitor",
  "agreement": true,
  "generation_type": "adversarial"
}
```

- `engine_result` — engine classification: "no_concern", "monitor", or "refer"
- `ground_truth` — AI clinical evaluator's expected label
- `agreement` — whether they match
- `generation_type` — "hand_verified" (Phase 1) or "adversarial" (Phase 2)
- `domains.*.flags` — specific developmental red flags detected

## License & attribution

- **Data:** CC BY-SA 4.0 — use for research, products, publications; attribute and share alike.
- **Engine:** Apache-2.0.
- **Citation:** Songra, H. & Ansari, A. (2026). "MyChild Engine Validation Study: Synthetic Developmental Screening Data." Available at www.mychildapp.in/study-data.

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# MyChild Engine — Documentation

> API reference, quickstart guide, and architecture documentation for the `mychild-engine` npm package. The docs page renders the repository README at build time; the always-current source is the GitHub repo.

- Canonical HTML page: https://www.mychildapp.in/docs
- GitHub (full README & docs): https://github.com/hsongra11/mychild-app
- npm: https://www.npmjs.com/package/mychild-engine (latest: 0.4.0, Apache-2.0)

## What the engine is

An open-source developmental screening engine for children 0-36 months with evidence-weighted escalation:

- **131 caregiver-facing questions** across 10 age bands (birth to 36 months), aligned with CDC 2022 revised milestones (Zubler et al., *Pediatrics* 2022, 75th percentile standard)
- **8 independent domains**: Gross Motor, Fine Motor, Receptive Language, Expressive Language, Social-Emotional, Cognitive, Self-Help/Adaptive, Vision/Hearing
- **Evidence-weighted escalation**: weight classes (Low/Medium/High/Red-Flag) actively adjust grace windows — high-weight items escalate faster, low-weight items get more grace
- **Corrected age** for preterm infants (born <37 weeks) until 24 months
- **Explainability**: every severity output includes a plain-English "why" trace
- **Rule simulator**: replay synthetic child timelines against different threshold settings

## Quickstart

```bash
npm install mychild-engine
```

```js
import { evaluate, computeChildAge, getDueQuestions } from 'mychild-engine';

const child = { dob: new Date('2025-09-01') };

// Get due questions for the child's (corrected) age
const questions = getDueQuestions(child, []);
console.log(`${questions.length} questions due`);

// Evaluate answers
const result = evaluate(child, [
  { questionId: questions[0].id, answer: 'achieved', timestamp: new Date() },
  { questionId: questions[1].id, answer: 'not_yet', timestamp: new Date() },
]);
// result includes per-domain severity + plain-English trace
```

## Validation CLI

```bash
npx mychild-engine validate --profiles data/synthetic-profiles.json --format markdown
npx mychild-engine autoresearch --runs 4 --adversarial 70000
```

## Licensing notes

- Engine code: Apache-2.0. Question bank & data: CC BY-SA 4.0.
- The question bank derives from publicly available CDC milestone checklists — it does not reproduce ASQ-3, M-CHAT-R/F, Denver, or other copyrighted instruments. Integrating those requires licenses from their rights holders.
- Not a diagnostic tool; not clinically validated. See https://www.mychildapp.in/terms.md

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# MyChild Engine — RBSK Alignment

> Domain-by-domain mapping between MyChild Engine's 131 questions and RBSK (Rashtriya Bal Swasthya Karyakram), India's national child health screening program. Alignment for integration planning — NOT endorsement, NOT RBSK validation.

Canonical HTML page: https://www.mychildapp.in/rbsk

RBSK screens children from birth to 18 years for developmental delays, defects, diseases, and deficiencies — roughly 270 million children across India. It has been independently validated at 97% sensitivity and 96.4% specificity against ASQ-3 (PPV 99.6%, Cohen's kappa 0.854; source: "Validation of RBSK screening tool… in Bengaluru", PMC 2024; an Assamese-version validation was published in 2025).

**To be clear:** mapping our questions to RBSK doesn't make this RBSK-validated. The scoring logic hasn't been tested against RBSK benchmarks or with Indian populations.

## Domain mapping

| MyChild Engine | RBSK equivalent | Overlap |
|---|---|---|
| Gross Motor | Motor development | Full — head control, sitting, standing, walking |
| Fine Motor | Motor development (fine) | Partial — RBSK covers basic grasp; we're more granular (pincer, transfer, stacking) |
| Receptive Language | Language / Communication | Partial — we separate receptive from expressive; RBSK combines them |
| Expressive Language | Language / Communication | Partial — same as above |
| Social-Emotional | Social / Personal | Good — social responsiveness, attachment, play |
| Cognitive / Play | Cognitive | Partial — we cover problem-solving and object permanence; RBSK is less granular |
| Self-Help / Adaptive | Personal / Adaptive | Partial — feeding, dressing, toileting covered by both |
| Vision / Hearing | Sensory screening | Partial — RBSK has separate sensory protocols; we embed sensory items within milestones |

## How the two systems differ

| Aspect | MyChild Engine | RBSK |
|---|---|---|
| Age range | 0-36 months | 0-6 years |
| Questions | 131 weighted questions | ~30 screening items (0-3 years) |
| Scoring | Evidence-weighted severity with grace windows | Pass/fail with referral threshold |
| Domains | 8 independent domains | 4 broad categories |
| Evidence weights | Low / Medium / High / Red-Flag | Binary |
| False positive reduction | 5 probe types (P1-P5) | Limited follow-up guidance |
| Regression detection | Automatic (achieved → not_yet) | Manual clinician judgment |
| Preterm correction | Automatic until 24 months | Clinician-directed |

## What each adds

**MyChild adds to RBSK:** evidence-weighted scoring with configurable thresholds; 5 probe types reducing false positives; automatic regression detection; corrected age for preemies; per-domain confidence with sufficiency gates; a rule simulator for testing threshold changes.

**RBSK has that we don't:** 3–6 year coverage; integration with India's health infrastructure (ABDM, Poshan Tracker); trained health worker protocols; referral pathways to District Early Intervention Centres (DEICs); screening beyond developmental delays.

## Integration opportunities

- **ABDM (Ayushman Bharat Digital Mission):** engine outputs as structured screening data in ABHA health records.
- **Poshan Tracker:** combined growth + developmental monitoring in the first 36 months.
- **RBSK Portal:** the engine as a digital scoring layer on RBSK's largely paper-based screening — same questions, smarter scoring.

## References

1. RBSK Portal, National Health Mission — https://rbsk.mohfw.gov.in/RBSK/
2. "Validation of RBSK screening tool in Screening of Developmental Delay among under 3-year children in Bengaluru." PMC, 2024.
3. "Validation of Assamese Version of RBSK Tool for Early Detection of Developmental Delay and Autism Spectrum Disorder." PubMed, 2025.
4. National Health Mission RBSK page — https://nhm.gov.in/index4.php?lang=1&level=0&linkid=499&lid=773
5. Full alignment document: https://github.com/hsongra11/mychild-app/blob/main/docs/research/rbsk-alignment.md

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# MyChild Engine — Privacy Policy

> The screening engine runs entirely on your device and transmits no child data anywhere. This website collects the minimum a website collects.

Canonical HTML page: https://www.mychildapp.in/privacy · Last updated: July 11, 2026

## Your child's data never leaves your device

The MyChild Engine is a local-first library. Every screening — questions, answers, scores, explanations — is computed on the device where it runs. The engine has no server component, no telemetry, and no network calls. We never see, store, or transmit any information about any child. Products built on top of the engine have their own data practices; check their privacy policies.

## What this website collects

- **Analytics:** Google Analytics (GA4) for page views and traffic. Sets cookies; shares usage data with Google. No account, no names, no child data.
- **Hosting logs:** served by Vercel, which keeps standard server logs (IP, user agent, request path) for operations and abuse prevention.
- **Study data requests:** if you request the raw study data, we store the email you enter and use it once — to send download links (via Resend, stored in Convex). No newsletter, no marketing list, no resale.
- **Embedded video:** the homepage embeds a YouTube video, which may set Google/YouTube cookies.

## What we don't do

- No selling or sharing personal data for marketing
- No data about children collected anywhere — the website has no child-related forms and the engine runs offline
- No ads or ad trackers
- No accounts or logins

## Your choices

Block analytics with any content blocker — the site works fine. To have a study-data request email deleted, contact us (https://www.mychildapp.in/contact.md). The study datasets contain only synthetic profiles — no real children.

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# MyChild Engine — Terms of Use

> Plain-language terms: licenses (Apache-2.0 code, CC BY-SA 4.0 data), medical disclaimer, no warranty.

Canonical HTML page: https://www.mychildapp.in/terms · Last updated: July 11, 2026

## Not medical advice. Not a diagnostic tool.

MyChild Engine and everything on this website are for informational and educational purposes only. The engine tracks developmental milestones and surfaces patterns; it cannot diagnose any medical condition, developmental disorder, or disability, and it is not a substitute for professional medical advice, diagnosis, or treatment. Its thresholds have been verified against synthetic scenarios but have NOT been validated in clinical trials with real children. Always consult a qualified healthcare professional about your child's development. If your child loses skills they previously had, seek medical advice promptly. In an emergency, contact your local emergency services.

## Licenses

- **Engine code:** Apache-2.0 — commercial use, modification, distribution allowed.
- **Question bank & study data:** CC BY-SA 4.0 — attribute, share derivatives alike.
- **Third-party instruments:** the question bank derives from publicly available CDC milestone checklists and does not reproduce ASQ-3, M-CHAT-R/F, Denver, or any other copyrighted instrument. Integrating validated instruments into a product built on this engine requires licenses from their rights holders.

## No warranty, limitation of liability

The engine, data, and website are provided "as is", without warranty of any kind, per the Apache-2.0 license. To the maximum extent permitted by law, the authors and contributors are not liable for any claim, damages, or other liability arising from use of the software, data, or website — including decisions made or not made about a child's health based on engine output. The engine's output is a signal to talk to a professional, never a decision.

## If you build on the engine

You're responsible for regulatory compliance in your jurisdiction, your own privacy practices, and presenting results with appropriate disclaimers. We ask that you keep the "not a diagnostic tool" framing and escalation guidance intact wherever screening results are shown to caregivers.

## Website content

Text on this website may be quoted with attribution and a link. Founder photos and third-party logos remain the property of their respective owners.

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)

---

# MyChild Engine — Contact

> Two people build this. You'll reach a founder, not a support queue.

Canonical HTML page: https://www.mychildapp.in/contact

## Bugs, questions, contributions

Anything about the engine, the question bank, the study, or this website — open a GitHub issue (public and searchable, so others benefit): https://github.com/hsongra11/mychild-app/issues

## Clinical partners

A prospective clinical study with real children is the next milestone. Clinicians, researchers, hospitals, and public health programs interested in validating the engine: message Harsh Songra on LinkedIn — https://linkedin.com/in/harshsongra

## Press & speaking

- Harsh Songra: https://harshsongra.com
- Aafreen Ansari: https://linkedin.com/in/aafreen-ansari-8a4265112

## Privacy requests

Requested study data and want your email deleted? Open a GitHub issue or message us on LinkedIn. See the privacy policy: https://www.mychildapp.in/privacy.md

---

Pages: [Home](/index.md) · [About](/about.md) · [Study](/study.md) · [Study Data](/study-data.md) · [Docs](/docs.md) · [RBSK](/rbsk.md) · [Privacy](/privacy.md) · [Terms](/terms.md) · [Contact](/contact.md)