Headline

Software Engineer @ Spire.AI

About

I'm a Data & Machine Learning Engineer who enjoys turning messy, unstructured data into systems that make decisions easier for people and businesses. Over the past year at Spire.AI, I've built a production-grade conversational AI onboarding platform using FastAPI, WebSockets, and AWS Bedrock (Claude), combining real-time voice AI (speech-to-text and text-to-speech) with LLM-driven data extraction — replacing traditional forms with natural conversation. I've also worked on skill taxonomy generation, embedding-based matching, and NLP model fine-tuning (T5, DeBERTa, SpaCy, LLaMA, Qwen, Gemma) at scale, cutting manual effort by 80%+ across several workflows. What I care about: → Turning unstructured text into structured, usable data through NLP and machine learning → Building reliable, scalable data pipelines — not just models that work in a notebook → MLOps practices that keep models honest in production (monitoring, drift detection, experiment tracking) → Clear, data-driven insights that support real business and workforce decisions Technical toolkit: Python, TensorFlow, PyTorch, Scikit-learn, LangChain, Transformers, AWS, Docker, Kubernetes, SQL, Power BI/Tableau. Currently exploring: large language model fine-tuning, agentic tool-use architectures, and applied NLP for workforce intelligence. Open to connecting with fellow ML/NLP engineers, data scientists, and anyone building with LLMs.

Location

Bengaluru

AI stack

claude code

claude

hugging face

langchain

langgraph

langsmith

Work experience

Software Engineer 1 at Spire.AI, Jan 2026 - Present · 10 mos, Bengaluru, Karnataka, India, - Built a production-grade conversational AI onboarding system using FastAPI, WebSockets, and AWS Bedrock (Claude), integrating real-time voice AI (Vertex AI STT, Nova Sonic TTS) for multi-modal user experiences. - Developed and deployed AI/ML pipelines — skill taxonomy generation, NLP model fine-tuning, and embedding-based matching — processing large-scale data and reducing manual effort by 80%+. - Analyzed large-scale job and workforce datasets to identify patterns in role composition and automation potential. Applied data decomposition techniques to break down job roles into granular, measurable task units. Conducted statistical analysis across 7,000+ skill clusters to evaluate automation exposure and impact. Translated analytical findings into actionable insights to support workforce planning and reskilling strategy decisions.

Data Analyst at Spire.AI, Apr 2025 - Dec 2025 · 9 mos, Bengaluru, Karnataka, India, - Designed automated web-scraping and parsing workflows (Octoparse, Selenium) for dynamic webpages, enabling reliable continuous data collection with minimal manual intervention. - Cleaned and matched large-scale certification datasets, improving accuracy to 93%+ through statistical validation. Automated manual data-matching tasks, reducing processing effort by 80%.

Junior data scientist at Solar secure IT Solutions, Dec 2023 - Apr 2024 · 5 mos, - Built a Convolutional Neural Network (CNN) in TensorFlow/Keras for automated product image classification. - Applied preprocessing and data augmentation techniques (resizing, normalization, augmentation) to improve model accuracy and generalization.

Data Analyst at Intern Carrer, Oct 2023 - Dec 2023 · 3 mos, Maharashtra, India

Education

Bachelor's degree, Computer Science, Vignan Institute of Technology and Management (VITAM), Brahmapur

Data scientist, Ekeeda - School of Data Science

Public links

github: https://github.com/Imajitdas7