Nishit Jain Email & Phone Number

Queens, New York, United States

Stony Brook, New York, United States

• Engineered a system for automating zero-shot code generation and repair using large language models, including ChatGPT and open-source instruct models like CodeLLama2 and Starcoder.• Designed custom signals to prompt large language models to refine and fix generated code, improving success rate on coding competition benchmarking datasets by 10%.

San Jose, California, United States

• Leveraged time-series analysis techniques to accurately forecast customer usage patterns and enhance user experience through smart recommendations, leading to a 20% increase in customer satisfaction scores.• Devised a novel metric in PySpark SQL to use within a clustering framework for customer segmentation, improving tailored recommendations and informing product development decisions.

Stony Brook, New York, United States

• Prepared and delivered lectures to 100+ students on NLP topics such as word embeddings (Word2Vec) and sentence embeddings (Deep Averaging Networks, Long Short-Term Memory (LSTM), Transformers).• Developed assignments in PyTorch and deployed using Amazon EC2 and Amazon S3 for the training, validation (per- plexity, BLEU score) and explainability (probing, de-biasing) of large language models.

Bengaluru, Karnataka, India

• Created synthetic NLP data and trained a multi-model classification pipeline using ensemble learning techniques from Scikit-Learn to auto-match B2B payments, improving precision by 15% and reducing manual workload by 32%.• Deployed a regression pipeline using Amazon SageMaker, Amazon S3 and Amazon EC2 to forecast invoice payment dates and generate future cash flow trends, achieving a 45% improvement in accuracy over baseline models.• Achievements: Received Certificate of Excellence under Spot Award category in Feb 2022 for ’valuable contribution to the multi-model classification architecture’.

Bengaluru, Karnataka, India

• Built an end-to-end pipeline in Python to automate sentiment classification of text documents using NLTK for data preprocessing, Gensim for feature extraction and LSTM for sequence modeling.• Developed robust time-series forecasting models to predict product revenue using historical sales data, resulting in a 25% reduction in the percentage error of prediction.• Led a team of 3 Data Scientists to develop real-time anomaly detection algorithms for high-frequency time-series data, identifying transient anomalies with a 60% precision.• Built an RMarkdown notebook to automate end-to-end analysis of multivariate time-series data using ARIMA, Vector AutoRegression and their variants, which enabled client-side teams to reduce model development time by 2 weeks.• Developed a Tableau dashboard summarizing KPIs such as Active Users and Frequency of Use for internal tools at Mu Sigma to assist C-suite executives in making critical decisions for business units.• Achievements: Received Certificate of Excellence under Spot Award category in Feb 2020, July 2020, Nov 2020, Feb 2021 and Apr 2021 for ‘driving the Algos thread and delivering the highest quality of work’.

Noida Area, India

• Built a multi-class classification model to identify the lease absorption potential of channel partners, thus decreasing the default rate by 25% for a financial services fir

Master Of Science - Ms, Computer Science, 3.96 / 4.0

Data Science and Engineering Specialization

Bachelor Of Engineering, Information Technology, 3.8 / 4.0