Debdipta Halder

About Debdipta Halder

FireCompass is a SaaS platform for Continuous Automated Red Teaming (CART) & Attack Surface Management (ASM). It continuously indexes and monitors the deep, dark & surface web to map out an organization\'s digital attack surface, including Shadow IT blind spots. The platform then automatically launches safe multi-stage attacks, mimicking a real attacker, to help identify attack paths before hackers do.Bridging the gap between cutting-edge AI research and scalable production engineering.With 9 years of total experience and the last 5 years dedicated to AI & Machine Learning, I operate at the intersection of theory and practice. My background as a Lead Software Engineer allows me to approach AI research not just as a scientific experiment, but as a product component waiting to be built.What I Bring to the Table:Applied Research: I don\'t stop at the paper. I specialize in researching Language Models (LLMs), fine-tuning them for domain-specific use cases, and optimizing them for real-world constraints.Engineering Rigor: My solutions are designed for production from Day 1. I build robust POCs and architectural blueprints that Engineering teams can seamlessly mitigate to production.Innovation & IP: Passionate about pushing boundaries, with 2 research papers pending publication and a patent currently in filing.Core Competencies:GenAI & LLMs: Fine-tuning (PeFT/LoRA), RAG Architectures, Agentic Workflows, Prompt Engineering.Development: Python, PyTorch, TensorFlow, MLOps pipelines. Golang, MCP Services.Leadership: Mentoring junior researchers and bridging communication between Data Science and Engineering teams.I am driven by the challenge of taking complex,\"impossible\" engineering problems and solving them through innovative AI strategies.

Experience

1. Ai Researcher

   Firecompass

   Aug 2025 — Present · IN

   Role: Lead AI Researcher Description:Leading the research and development of domain-specific Large Language Models (LLMs), focusing on fine-tuning and alignment (RLHF/DPO) to align with product goals.Research-to-Production Pipeline: Successfully architected and delivered [Number] high-impact POCs, creating the technical blueprint for the Engineering team to scale to production.Solving complex engineering bottlenecks using AI-driven optimization, reducing [metric, e.g, latency/cost/manual effort] by [X]%.Filed a patent for a Novel Embedding Solution for the Cybersecurity Domain, introducing a novel approach to transforming complex threat telemetry into high-fidelity vector representations to enhance automated detection and response. Technical Overview:[Status]: Patent Pending[Area]: Machine Learning / Cybersecurity[Focus]: Advancing the state of the art in how we represent and analyze security-relevant data.Authored 2 research papers on \"NOVEL EMBEDDING SPACE FOR CYBERSECURITY\" and \"TAG STANDARDIZATION USING LLM\", contributing to the organization\'s thought leadership in Generative AI.Mentoring a team of researchers and engineers on best practices for MLOps and model evaluation.

Education

• NISER

  Master's Degree (5 years Integrated), Mathematics

  2008 — 2013

• freeCodeCamp

  Full Stack Web Development Certification, Computer Software Engineering

  2017 — 2018

• Indian Institute of Technology, Kharagpur

  M.Tech, Computer Science

  2013 — 2015

Skills

• Xslt
• Data Structures
• Microsoft Office
• Microsoft Word
• Jquery
• Shell Scripting
• Mysql
• Oracle Sql Developer
• Pl/Sql
• C
• Programming
• Research
• Matlab
• Sql
• Linux
• Software Development
• Python
• Bootstrap
• Microsoft Excel
• Html
• Oracle Database
• Java
• Machine Learning
• Algorithms
• Javascript
• C++