Jul 2025 – Sep 2025 · 2 mos · Boston, Massachusetts, United States · Remote

Jan 2025 – May 2025 · 4 mos · Boston, Massachusetts, United States · Remote

• Accomplishments include:
• > Optimized training for clinical NLP by designing a graph-based data sampling pipeline that improved F1 scores by 20%—cutting down annotation fatigue and boosting model generalization.
• > Fine-tuned LLaMA-3.1-8B and Phi-2 for relation extraction on clinician notes, incorporating expert feedback from medical professionals into the fine-tuning loop to improve interpretability, domain relevance, and compliance with clinical standards.
• > Deployed models on AWS SageMaker, supporting real-time inference and scalable multi-model management in production using LoRA adapter switching—reduced deployment overhead by 40%.
• > Improved prompt reliability by identifying sources of hallucination and ambiguity, and redesigned prompt templates to raise precision by 15% across tasks.
• > Collaborated across ML, product, and clinical teams to ensure that AI outputs were not only technically sound but actually trusted and usable by domain experts.

May 2024 – Aug 2024 · 3 mos · Hoboken, New Jersey, United States · Remote

• Accomplishments include:
• > Researched and implemented prompt engineering techniques using open-source LLMs (LLAMA3) to improve the model’s accuracy in detecting 23 persuasion techniques, resulting in a 25% increase in classification accuracy.
• > Conducted extensive performance tuning and experimentation with split prompts, from small splits (3 techniques per prompt) to larger splits (all 23 techniques in one prompt), optimizing the model’s efficiency and improving its ability to handle complex classification challenges.
• > Devised a method for integrating syntactical information in prompts by grouping persuasion techniques based on rhetorical and grammatical patterns, which enhanced the LLM’s retention of grammatical data, validated through fuzzy logic analysis.
• > Applied self-consistency across various prompt configurations to boost the reliability and performance of the LLM in multi-class classification tasks, achieving a 20% performance improvement.

Feb 2023 – Aug 2023 · 6 mos · Gurugram, Haryana, India · On-site

• Accomplishments Include:
• >Devised an autonomous Python-based utility for reporting changes in regulatory reports, elevating data analysis and reporting processes by an impactful 60%.
• >Optimized data flows by partitioning big datasets using relevant features in Spark, resulting in a 40% improvement in data processing speed and a 25% enhancement in query performance through SQL servers and queries
• >Tracked and resolved 20 critical bugs in custom Python plugins for an in-house data management tool, through meticulous testing.
• >Mentored interns, fostering knowledge transfer and task delegation that resulted in a decrease in learning curves for the interns.

May 2022 – Aug 2022 · 3 mos

• Accomplishments Include:
• > Implemented the development and integration of a Python-based post-processing module, rectifying punctuational and syntactical errors in a custom Speech Recognition model, yielding a substantial 20% improvement in overall accuracy.
• > Designed an innovative feature selection technique for an object detection model, reducing processing time by 20% while maintaining high accuracy rates.
• > Analysis and tagging of 1000 data points to create a high-precision Named Entity Recognition (NER) model enhanced overall the data quality.

May 2021 – Jul 2021 · 2 mos · Mumbai, Maharashtra, India · Remote

• Accomplishments include:
• > Collaborated seamlessly with senior developers to elevate the performance of client-based applications, leading to a notable reduction in user interface errors and enhancing overall user satisfaction.
• > Implemented advanced coding techniques and practices recommended by senior mentors, to improve application performance and decrease in system crashes.