Curriculum vitae

pdf

Mohit Kumar

Bengaluru, India
mohit.iit16@gmail.com


NLP | Generative AI | Full-Stack LLM Development

# Summary

Research Engineer with 7+ years of experience designing, implementing, and optimizing AI solutions at scale. Expertise in machine learning, deep learning, and NLP, with a strong focus on Large Language Models (LLMs) such as GPT, Llama, and BERT. Proficient in ML frameworks (PyTorch, TensorFlow, scikit-learn) and deployment using Docker, and cloud platforms (AWS, GCP). Strong problem-solving and collaboration skills, with experience delivering high-impact AI solutions.

# Skills:

  • Data Science/Machine learning: Pandas, sklearn, Keras, PyTorch
  • Deployment & Cloud: Docker, AWS, GCP
  • LLM & Prompt Engineering: Prompt Optimization, LLM Inferencing, OpenAI GPT-4, Anthropic Claude, LangChain, PromptLayer, Guidance
  • Web Frameworks: Django, Flask, FastAPI, Jinja2, Gunicorn, Uvicorn
  • ORMs & Database Tools: SQLAlchemy, Django ORM, Alembic, Postgres
  • Concepts: Data Structures and Algorithms, Natural Language Processing, Deep learning, GenAI, RAG

# Education

Bachelor of Technology, Indian Institute of Technology (Banaras Hindu University), Varanasi

# Employment History

Freelance Developer - Tinybyte.ai (May 2025 - Present)

  • TBA

Senior Research Engineer - American Express (Apr 2022 - Apr 2025)

  • Vocal - Voice of Customers: Spearheaded the implementation of a new ETL pipeline, streamlining the data collection process for customer complaints and enhancing overall data integrity for advanced analysis. Developed a Retrieval Augmented Generation (RAG) pipeline utilizing advanced Large Language Models (LLMs), significantly improving the accuracy of analysis and reducing human efforts by more than 80%. Collaborated with cross-functional teams to extract insights from specific datasets, delivering actionable recommendations with rapid turnaround times to support decision-making.
  • Model Interpretability for Large Language Models (LLMs): Developed and implemented model interpretability guidelines to ensure transparency and trustworthiness of GPT-based Large Language Models (LLMs) for enterprise-wide applications. Designed and introduced a novel evaluation mechanism to assess the performance, reliability, and scalability of LLMs, enabling robust deployment of LLM-based solutions across the organization.
  • TTYD - Talk to Your Data: Responsible for developing a question answering system that could process natural language queries and deliver fast, accurate (upto 87%) responses to users. Deployed it into production, monitored how it was performing, identified ways to improve response times and latency, and implemented optimizations to achieve a better overall user experience.

Senior Data Scientist - EXL Service (Jun 2019 - Aug 2021)

  • Key-Concept Extraction: Developed and published Key-Concept Extraction Engine, an unsupervised model to automatically extract topics from call transcripts, served with a user-friendly web-app with backend based on Django and PostgreSQL. Employed Dynamic topic modeling (DTM) techniques to analyze the evolution of topics of conversation over time. Achieved 85% satisfaction score on manual validation of transcripts.

Machine Learning Engineer - Tvarit GmbH (Aug 2018 - Dec 2018)

  • Automated Predictive Analytics Tool (APA): Involved in development of APA tool, a Time Series Predictive Analytics tool which automatically does anomaly detection, feature selection, model tuning, and deployment as well as performance monitoring and logging.

# Languages

English (Fluent), Hindi (Fluent)