Montclair State University — Technical Analyst, Center for Academic Success and Tutoring

Montclair, NJ · Feb 2024 – Dec 2025

• Generated and maintained operational datasets across teams for 25,000+ student records; built dashboards and reports for real-time monitoring of engagement and program health.
• Architected fault-tolerant data pipelines integrating Tutor.com, study hall, and recitation data with schema evolution, freshness SLAs (<24h), and automated backfill; sustained <1.5% failure rate and reduced cross-system inconsistencies ~30% across 10K+ weekly records.
• Implemented validation and drift monitoring (PSI) with weekly cycles and alerts (PSI > 0.2); held >95% precision and ~85% recall on anomalies, cutting invalid records ~20%.
• Built an early-warning model (logistic regression, engineered features, class weights) with threshold 0.7 for recall under ≥0.65 precision; reached 0.81 ROC-AUC vs. 0.63 baseline, with rule-based fallback for low-confidence bands.
• Delivered retention modeling with XGBoost and SHAP on cohorts; improved F1 ~17% vs. heuristics and supported interventions that raised retention ~13% in high-risk segments.
• Quantified dropout drivers with confidence intervals and effect sizes; thresholds adopted by stakeholders, reducing false positives in targeting ~12%.

Accenture — Software Engineer (Full Stack)

India · Feb 2022 – Aug 2023

• Fine-tuned BERT (Hugging Face) for intent and NER on ~15K multilingual fintech queries; raised precision ~18% (with regional recall tradeoffs).
• Built locale-aware preprocessing (Unicode normalization, rules, regex) across 6+ countries; cut parsing failures ~35%.
• Developed ETL with Pandas/SQLAlchemy on ~8K–10K daily records; resolved schema/null/duplicate issues; cut ingestion errors ~40%.
• Exposed inference via Flask REST APIs (~20–25 req/sec, ~150–200 DAU); improved latency to ~280–320ms via tracing, connection reuse, and payload tuning.
• Strengthened Jenkins pipelines (build → validation → deploy) with schema gates; cut deployment breakages ~20% on multi-region releases.
• Also supported large-scale platforms: SQL/PostgreSQL, Kibana, incident response (200+ critical incidents), and performance work contributing to reliability at scale.

Infor — Junior Software Engineer

Hyderabad, India · Mar 2021 – Jan 2022

• Refactored Spring Boot endpoints using composite BTREE-indexed queries and pagination; cut latency ~480ms → ~310ms under ~20 concurrent users.
• Built an OCR microservice (Python, Tesseract, logistic regression features); precision ~0.68 → ~0.81 on ~1.2K labeled samples.
• Enforced integrity via JSON schema, type/null checks, and uniqueness; cut duplicates ~25% and stabilized transaction success >97% on ~8K weekly entries.
• Improved tests (JUnit/Mockito) and reduced GC pressure ~12% via profiling and DTO refactors; standardized Jenkins configs and cut deployment failures ~20%.