experiences

Sept 2025 – Present

Commure + Athelas

Operations Analyst (RCM Rules Engine) | Onsite
  • Designed SQL-based validation pipelines for the RCM Rules Engine by hitting LLM endpoints with structured rule JSON payloads via chain-of-thought prompting, enabling context-aware claim adjudication across large-scale EHR datasets.
  • Engineered an LLM-driven rule audit system using LangChain + GPT-4/Gemini with context caching over large payer logic schemas and historical denial records, where the model surfaces rule fix suggestions from exceptions, cutting manual review by 80%.
  • Built an internal RAG-powered AI agent using ChromaDB + pgvector embeddings over the backend codebase for context-aware retrieval of rule conditions, payer logic, and EHR schema, significantly reducing rule debugging and onboarding time.
  • Fine-tuned an in-house open-source LLM using QLoRA and a custom PyTorch training loop on sensitive claims denial payloads, improving model precision on payer-specific adjudication edge cases over the base model.
  • Built iterative analytics dashboards surfacing LLM-generated diagnostic insights on rule coverage, denial rate trends, and exception clustering across production claim workflows.

2023

Qualcomm

Software Engineer Intern (QDSP Team) | Hyderabad, India
  • Enhanced power measurement accuracy for SNPE and QNN SDKs, leading to a 30% reduction in energy consumption for ML models.
  • Developed a robust algorithm to measure and visualize Inference/watt for ML models, increasing analysis efficiency by 40%.
  • Identified and resolved critical bugs, improving overall system reliability by 25%.
  • Authored comprehensive documentation that improved team efficiency by 20% and reduced onboarding time for new hires by 50%.

2023 - 2024

National Institute of Technology, Rourkela

Research Assistant
  • Collaborated with Prof. Tapas Kumar Mishra on advancing machine translation for low-resource Indic languages.
  • Integrated linguistic features (POS tagging, NER, lemmatization) into Transformer models, improving translation quality by 35%.
  • Trained models using OpenNMT-py on the Samanantar dataset to achieve state-of-the-art results.

2021 - 2023

Chittagong Junior Coders

Instructor | Online (Bangladesh)
  • Taught C programming to students from 2nd to 10th grade, making coding accessible and enjoyable for young learners.
  • Fostered a strong foundation in coding principles and problem-solving skills, igniting a passion for technology.
  • Volunteered time and knowledge to create a supportive learning environment for the next generation of programmers.