NLP researcher turned ML systems engineer.

If it's worth doing, it gets an architecture, a README, and too much Python 🦸👊

Started in NLP and computational linguistics, ended up owning the full lifecycle — data pipelines, model training, deployment, monitoring, and the platform underneath it all.

I gravitate toward problems at the intersection of language and scale: multilingual text understanding, embedding-based retrieval, and the kind of messy real-world data that makes clean solutions fall apart.

Currently at deecoob GmbH (Dresden) — entity resolution, multilingual classification, and ML infrastructure serving production traffic across DE/PL/NL.

📩 sachnandmenon@gmail.com · LinkedIn · Resume

linguaalayam — RAG pipeline over a 58k-entry Malayalam dictionary. Hybrid retrieval: exact match, trigram fuzzy, HNSW cosine. Exposed as an MCP server. LangGraph pgvector LiteLLM FastMCP cross-encoder reranking

readme-drift — CLI + pre-commit hook for stale README detection. AST-based diff analysis catches renamed symbols, removed functions, shifted paths. Published to PyPI. Python AST pre-commit GitHub Actions PyPI

job-ledger — End-to-end job intelligence pipeline. Scrapes 200+ listings/run, Claude-powered CV-to-JD scoring, Streamlit UI, SQLite persistence. Apify Anthropic SDK Streamlit SQLAlchemy

• NLP & ML — Python · HuggingFace Transformers · SentenceTransformers · PyTorch · XGBoost · SpaCy
• LLM & retrieval — LangGraph · LiteLLM · FastMCP · pgvector · Elasticsearch
• MLOps & infra — Docker · Kubernetes · GCP · Terraform · ArgoCD · Kafka · GitHub Actions · Bitbucket Pipelines
• APIs & data — FastAPI · gRPC · PostgreSQL · SQLAlchemy · Alembic