Bhavya
Request intro
// PRODUCTION DEPLOYMENTS & ARCHITECTURAL CASE STUDIES · SECTION 04

Flagship Systems & Production Impact

Technical architecture breakdowns of production systems, projects, and repositories this candidate has attached to their dossier.

// INTERVIEW ENGINE

Interview Engine is arriving soon....

SYSTEM 01 // PROJECTRole: FEATURED PROJECT

Autonomous Lead Enrichment Agent

Built an autonomous B2B lead-enrichment agent that crawls company domains with a headless browser and extracts structured firmographic data via Gemini Flash, validated with Pydantic schemas. Features async scraping with asyncio.Semaphore concurrency control, temperature=0.0, per-record confidence scores, and a success/partial/failed error taxonomy; tested on 3 sites (0.85-0.95 confidence) with an argparse CLI.

TechnologyPython
StackPlaywright
PlatformGoogle Gemini Flash
Outcometested on 3 sites (0.85-0.95 confidence)Self-reported
Metric 02—Pending
Metric 03—Pending
View Project

// INTERVIEW ENGINE

Interview Engine is arriving soon....

SYSTEM 02 // PROJECTRole: FEATURED PROJECT

The Committee Will Look Into It

Building a 5-thread data journalism project on Indian institutional accountability (3 complete), using hand-built datasets, NCRB data, and RTI-sourced documents; found 110 exam-paper leaks led to only 8 convictions, with an 8.5-year average time to justice. Analysed 22 years of NCRB data on crimes against women and 99 internet shutdowns (Sep 2024-Jul 2026): showed a 21.3% national conviction rate can hide 98% case pendency (Bihar), and that 64% of shutdowns were preventive, with no triggering event.

TechnologyPython
StackPandas
PlatformMatplotlib
Outcomefound 110 exam-paper leaks led to only 8 convictions, with an 8.5-year average time to justiceSelf-reported
Metric 02—Pending
Metric 03—Pending
View Project

// INTERVIEW ENGINE

Interview Engine is arriving soon....

SYSTEM 03 // PROJECTRole: FEATURED PROJECT

Message Spam Filtering

Built and deployed an NLP classifier on 5,572 SMS messages; switched from MultinomialNB to ComplementNB to address 87/13 class imbalance, improving spam recall from 72% to 89% at 97.4% test accuracy. Engineered bigram TF-IDF features (ngram_range=(1,2)) to capture phrase-level spam patterns missed by unigrams; applied stratified train/test split with full-data refit before deployment.

TechnologyPython
StackScikit-learn
PlatformTF-IDF
Outcomeimproving spam recall from 72% to 89% at 97.4% test accuracySelf-reported
Metric 02—Pending
Metric 03—Pending
View Project

// INTERVIEW ENGINE

Interview Engine is arriving soon....

SYSTEM 04 // PROJECTRole: FEATURED PROJECT

Customer Churn Prediction

Built classification pipeline on telecom churn data; benchmarked Logistic Regression, Random Forest, and SVM under K-Fold CV with 80/20 class imbalance handled via class_weight='balanced' on Logistic Regression. Benchmarked best model at 81% churn recall vs 0% for a naive majority-class baseline scoring 83% accuracy — framed recall, not accuracy, as the metric that actually matters for retention decisions.

TechnologyPython
StackScikit-learn
PlatformStandardScaler
Outcomebenchmarked best model at 81% churn recall vs 0% for a naive majority-class baseline scoring 83% accuracySelf-reported
Metric 02—Pending
Metric 03—Pending
View Project

// INTERVIEW ENGINE

Interview Engine is arriving soon....

SYSTEM 05 // PROJECTRole: FEATURED PROJECT

Barbie – Voice-Activated AI Assistant

Built real-time voice AI pipeline bridging Google Speech Recognition (ASR) → Groq LLaMA 3.3 70B inference → gTTS/Pygame audio output, achieving sub-3s round-trip interaction latency with autonomous web control. Integrated live NewsAPI headlines, YouTube playback, and custom prompts for a distinct AI personality with modular, extensible skill design.

TechnologyGroq LLaMA 3.3 70B
StackgTTS
PlatformNewsAPI
Outcomeachieving sub-3s round-trip interaction latencySelf-reported
Metric 02—Pending
Metric 03—Pending
View Project

// INTERVIEW ENGINE

Interview Engine is arriving soon....

SYSTEM 06 // REPOSITORYRole: OPEN SOURCE

The-Committee-Will-Look-Into-It

Multi-domain EDA and data storytelling project analyzing institutional accountability in India across 5 modules — NLP, time-series, and statistical analysis on government-published datasets (NCRB, SBI/ECI, ADR). Python | Pandas | Matplotlib | Seaborn. In Progress.

LanguageJupyter Notebook
Topic—
Topic—
Metric 01—Pending
Metric 02—Pending
Metric 03—Pending
View Repository

// INTERVIEW ENGINE

Interview Engine is arriving soon....

// INTERVIEW ENGINE

Interview Engine is arriving soon....

// RECRUITMENT & ADVISORY ENGAGEMENTS

Request Technical Architecture Dossier

Full verification packet containing source artifacts, production traces, and formal proofs available for prospective leadership roles once attested.

Download PDF (Signed)Request Intro

Hiring for Data Scientist?

Request the full evidence dossier — verified interview sessions, cryptographically attested artifacts, and reference outcomes.

Request intro

Goes straight to Bhavya. No recruiter spam.

Download résumé
Verification key —