Udisha Dutta Chowdhury
Machine Learning Engineer @Squark Ai
Signup · Get unlimited contacts
WORK HISTORY
Machine Learning Engineer @Squark Ai
US
Delivered production AI engineering across 2 organizations, Humanitarians AI and Squark AI• Architected multi-agent AI system with LLM orchestration via MCP protocol using Claude (Anthropic) — integrated LSTM/GRU forecasting with LLM advisory layer; deployed on AWS ECS with auto-scaling and CloudWatch observability• Built LangChain RAG pipeline with FAISS vector retrieval reducing query latency by 43%; engineered SHAP/LIMEexplainability modules; designed LLM evaluation framework using RAGAS — assessed faithfulness, answer relevance, context precision, and context recall• Restored ML feature engineering pipeline via DuckDB/Polars type inference fix for enterprise AutoML platform with Snowflake ETL workflows; implemented CI/CD with GitHub Actions and Docker containerization
EDUCATION
Northeastern University
Master's degree, Computer Systems Engineering (GPA: 4.0)
PES University
Bachelor of Technology - BTech, Electronics and Communications Engineering | Minor: Computer Science
ABOUT UDISHA DUTTA CHOWDHURY
I\'m an AI/ML Engineer specializing in agentic AI systems, LLM orchestration, and production machine learning — with a systems engineering foundation spanning IoT infrastructure and security engineering.My recent work has been at the cutting edge of applied AI: architecting multi-agent systems with LLM orchestration via MCP protocol, building LangChain-based RAG pipelines with hybrid retrieval (FAISS + BM25) that reduced query latency by 43%, and deploying production inference services on AWS ECS with auto-scaling and CloudWatch observability. I\'m drawn to the full stack of AI engineering — from agent reasoning and LLM integration to the data pipelines and infrastructure that make these systems reliable in production.That systems-first mindset comes from how I got here. I started my career at Deloitte building security detection infrastructure at scale — engineering YARA-L detection rules, developing Python automation frameworks that cut manual intervention by 40%, and designing threat detection pipelines processing millions of daily events. Working where data quality, latency, and adversarial conditions are non-negotiable gave me an engineering foundation that directly shapes how I build AI systems today.I brought that rigor to my Master\'s in Computer Systems Engineering at Northeastern University (GPA: 4.0), applying ML to industrial infrastructure at Schneider Electric — building XGBoost/LSTM ensembles for equipment failure prediction, engineering 40+ features from industrial sensor telemetry, and designing production ETL pipelines processing 500K+ daily time-series records.I also share this work publicly — presenting time-series forecasting research at PyData NYC and PyData Vermont, and submitting a proposal to SciPy 2025 for Backbencher, my open-source Python backtesting framework.What I\'m looking for: AI Engineer, ML Engineer, or Applied Scientist roles — in agentic AI, industrial AI, IoT systems, LLM-powered products, or security-adjacent ML. Open to all US locations.If you\'re building something at the intersection of AI and complex systems — I\'d love to connect. d••••••••@northeastern.edu
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.