Abjasree

1. 

   DATYCS

   lead data scientist · remote

   in productionApr 2024 — present

   • Led a 6-member GenAI & Research squad; built AWS/Azure CI/CD reducing model roll-out time by −80% (3 weeks → 5 days).
   • Architected & deployed 3 transformer-based NLP/CV pipelines processing 10K+ AML/CLL charts into Snowflake, boosting throughput +35%.
   • Fine-tuned TrOCR for handwritten prescriptions, improving accuracy +25% and enabling real-time medication safety alerts.
   • Deployed Vision LLM on Azure to parse multi-handwriting prior-authorization forms into structured JSON (+28% accuracy, −45% manual review).
   • Leveraged autonomous AI agents for patient-specific search validation and clinician response generation (+30% accuracy, −40% reply time).
   • Partnered with Eversana to analyze AML & CLL therapy-switch drivers and adverse effects, shaping pharmaceutical market-access strategies.
2. 

   IIT Madras

   data scientist · chennai

   archivedAug 2022 — Mar 2024

   • Automated analysis of 1,000+ CT/MRI images, reducing manual preprocessing time by −80%.
   • Increased renal calculi & pulmonary nodule diagnostic accuracy by +18% using CNN + radiomics + Bayesian-optimized XGBoost.
   • Deployed QA models at Apollo Proton Cancer Center, cutting turnaround −10% and flagging 12 high-risk plans per quarter.
3. 

   Univ.AI

   teaching assistant · part-time

   archivedDec 2021 — Aug 2022

   • Assisted in teaching machine learning and data science courses to students from diverse backgrounds.
   • Mentored students in practical ML projects and helped them understand complex AI/ML concepts.
   • Contributed to curriculum development and course material preparation.

1. [ pub ]

   Radiomics for NSCLC Classification ↗

   Oct 2023 — Jan 2024

   Built radiomics model with 83% accuracy for differentiating adenocarcinoma and squamous cell carcinoma in non-small cell lung cancer beyond nodule morphology in chest CT. Decision-curve analysis showed net clinical benefit across thresholds.

   RadiomicsMachine LearningMedical Imaging
2. [ conf ]

   Zero-Shot Image Registration (ZSIR-FE) ↗

   Sept 2022 — Oct 2023 · Med-NeurIPS 2023

   Designed DNN feature-based registration improving Dice score by +0.30 and running 6× faster than SIFT on BRaTs dataset through feature extraction approach.

   Deep LearningComputer VisionMedical Image Registration
3. [ pub ]

   Pulmonary Nodules Classification ↗

   Sept 2022 — Oct 2023

   Engineered radiomic features + XGBoost on limited Indian dataset, achieving 89% accuracy (+10% vs prior studies) for pulmonary nodules classification from chest CT scans.

   XGBoostFeature EngineeringMedical AI
4. [ lab ]

   Automatic Detection of Renal Calculi

   Sept 2022 — Jul 2023

   Processed 163 CT scans and built deep-learning pipeline reaching 85% accuracy (+25 pp improvement); slated for clinical deployment.

   Deep LearningMedical ImagingClinical AI
5. [ thesis ]

   Elliptic Flow of Light Nuclei ↗

   May 2019 — June 2020

   Simulated heavy-ion collisions with AMPT; computed elliptic flow using STAR detector data for light nuclei. Developed C++ macros to compare Coalescence models with experimental measurements, achieving thesis distinction. Also worked on simple ML classification models to distinguish signals and background.

   HPCC++Machine Learning

integrity checks — certifications

boot sequence — education

[ ok ] stage_0 · 2012 — 2013

Jawahar Navodaya Vidyalaya (JNV)

High School · CGPA: 10

[ ok ] stage_1 · 2013 — 2015

G. H. S. S. Palakkad

Higher Secondary · 97.33%

[ ok ] stage_2 · 2015 — 2020

IISER Tirupati

BS MS Dual Degree · CGPA: 8.0