SANKALP J.
INTRO
EXPERIENCE
RESEARCH
EDUCATION
CERTIFICATES
CONTACT
P.01
P.02
P.03
P.04
P.05
P.06
SANKALP JAJEE
Applied ML / Gen-AI Engineer
F-1 STEM OPT-eligible (3 yrs)
Seeking MLE / AI-research role

2018-2025

/02 Experience 

Languages

  • Python
  • SQL
  • C

ML / DL

  • PyTorch
  • TensorFlow
  • JAX
  • HuggingFace
  • MLflow

HPC / Accelerated

  • CUDA
  • NVIDIA V100 / A100
  • Slurm
  • NCCL
  • MPI
  • Horovod
  • DDP

Data / Ops

  • Spark
  • Docker
  • AWS
  • Git
  • Tableau

Awards

/01 Intro


My name is Sankalp K. Jajee, and I completed my Master’s in Computer Science at the University of South Carolina, specializing in Artificial Intelligence. My research spans in GenAI (multilingual evaluation benchmarks), recommender systems. I have created large-scale datasets such as IndicMMLU-Pro (108k QA pairs across nine Indic languages for ~1.7B speakers) and the Hierarchical Prompting Taxonomy (HPT), both of which are actively being adopted by the open-source and academic communities.


I have published multiple papers, including survey and benchmark contributions, Alongside my academic track, I bring five years of experience in building high-frequency trading pipelines, where I developed strong skills in AIML & Research

Sankalp Jajee

  • 01

    Medical University of South Carolina – Staff Scientist (ML)

    • Developed an AI chatbot to translate complex medical consent forms into lay-friendly language, improving patient comprehension and adherence to ethical guidelines.
    • Engineered an LLM pipeline to automate qualitative data coding from psychiatric transcripts, reducing manual analysis time by ~75% and improving coding consistency.
    • Built an adaptive learning AI chatbot for medical education with NetID authentication and a framework for personalized user guidance.
    • Developed a conversational AI (RAG) interface for the CIDER cancer data warehouse, enabling natural language queries across large-scale clinical and genomic datasets to accelerate research.
    Medical University of South Carolina | Charleston, USA
    Sept 2025 - Present

  • 02

    Graduate Researcher - Artificial Intelligence Institute of South Carolina (AIISC)

    • Fine-tuned 7B-parameter LLMs on 8×NVIDIA V100s using PyTorch and HuggingFace for multilingual tasks.
    • First-author of 4+ Gen-AI papers, including IndicMMLU-Pro (108k-row multilingual benchmark for ~1.7B speakers) and the Hierarchical Prompting Taxonomy (HPT).
    • Led projects on bias mitigation in text-to-video models, causal evaluation pipelines, and automated vigilance-state classification in rodents using ML.
    • Mentored graduate and undergraduate students (UC Davis, DTU) on applied ML research.
    University of South Carolina | South Carolina
    May 2023 – May 2025

  • 03

    UST Global - Software Developer (ML)

    • Re-engineered a Kafka → Spark → Grafana/XGBoost streaming pipeline (20k TPS), cutting P95 latency by 38 ms and reducing late-delivery rate by 12%.
    • Migrated ETL workflows from on-prem Hadoop → AWS EMR, reducing runtime by 38% and saving $42 k/year.
    • Introduced MLflow + GitHub Actions CI/CD, shortening the model release cycle from 10 days → 2 days.
    • Developed a synthetic-data generator (Faker + CTGAN), improving minority-class F1 score from 0.71 → 0.83.
    Client: Applied Materials (AMAT) | Bengaluru
    Jan 2022 – May 2023

  • 04

    Accrete Hi-Tech Solutions - Software Developer (ML)

    • Built a distributed cost-analysis platform using Spark (Adaptive Query Execution), cutting cluster-hours by 28%.
    • Designed 3D cost heatmaps (D3.js/WebGL), reducing analyst drill-down time from 15 min → 3 min.
    • Integrated Kafka Streams for anomaly detection at 40k msgs/sec, replacing an 8-hour lag with near real-time alerts.
    • Ran A/B model evaluation (GradientBoost vs. RandomForest) across 6 KPIs, lowering MAE by 19%.
    Client: Applied Materials (AMAT) | Bengaluru
    May 2021 – Dec 2021

  • 05

    QuantumLeap AI - Software Developer (ML)

    • Trained RL models on 4×NVIDIA V100 nodes via Slurm DDP, reducing epoch times by 62% and cutting AWS costs.
    • Built a Hybrid Prophet + LSTM model for stock prediction, improving recall by +35pp.
    • Applied Bayesian hyperparameter tuning (Optuna), lifting LSTM profit factor from 1.26 → 1.38 in just 7 GPU-hours.
    • Deployed ML models on AWS EC2/SageMaker, cutting inference latency by 27% with 99.9% uptime.
    • Leveraged Dynamic Time Warping (DTW) to compare stock patterns across markets, enabling pair-trading strategies.
    Software Development & Machine Learning | Bengaluru
    June 2018 – May 2021

/03 Research  

/04 Education 

MASTERS

Master of Science in Computer Science

/ University of South Carolina | USA

AUG 2023 - MAY 2025

BACHELORS

Bachelor of Engineering in Information Science and Engineering

/B.M.S College of Engineering | Bangalore, India

Aug 2014 – May 2018

/05 Certifications 

machine learning

Supervised Machine Learning

/Stanford University

2025

learning Algorithms

Advanced Learning Algorithms

/Stanford University

2025

LLMS

Fine Tuning LLM

/ DeepLearning.AI

2024

Recommender Systems

Unsupervised Learning Recommenders & Reinforcement Learning

/ Stanford University

2024

Collaborate

email me

linkedin

phone

/contact me

Know More

SEE MY RESUMé

GITHUB

GOOGLE SCHOLAR

/contact me

SANKALP JAJEE

SANKALP  J.

/25

email/

linkedin /

GOOGLE SCHOLAR /

github /

SANKALP'S SPACE

LET'S WORK ON EXCITING PROJECTS TOGETHER!

This website does not use cookies of any kind

© 2025