Ananya Pathak

Solopreneur and machine learning engineer building production systems at the intersection of healthcare, e-commerce, and SaaS.

Blog Email Book a call

experience

2022 – present

Pathak Labs patenting · e-commerce analytics · SaaS

Partnering with American and European startups on low-latency task-specific systems: semantic search, RAG, product embeddings, information extraction, image & text classifiers, and reinforcement learning in games. Evaluated AI feasibility for patent applications.

2018 – 2022

Machine Learning Researcher TERI · IGIB · IGBMC

Statistical and machine learning models in healthcare and the biological sciences — genomics, proteomics, and related fields.

now

Building PatViz · DataEngUtils · esmc.cpp

PatViz automates patent drawings for researchers and agents. DataEngUtils is a platform of data engineering utilities. I maintain esmc.cpp, a C++ inference engine for protein language models.

bio

I am a biomedical engineer turned machine learning (ML) engineer. I have worked in academic labs at TERI, IGIB, and IGBMC, building statistical and ML models across healthcare and the biological sciences.

As an independent consultant I have worked in the patenting, e-commerce analytics, and SaaS industries. Over the last couple of years I have been learning full-stack engineering and self-hosting software.

what i do

LLM fine-tuning & deep learning: supervised and RL fine-tuning, custom model training.
DataOps & MLOps: custom data and ML pipelines to terabyte (TB) scale and beyond.
AI agentic systems: autonomous, task-performing systems.
AI SaaS: backend, ML, frontend, CI/CD, deployment, scaling.
Hardware-efficient bio/large models: efficient architectures for biological sequence understanding.

Stack: PyTorch · FastAPI · Postgres · Docker · AWS · Hetzner · HTMX · and more.

projects

PatViz — automating patent drawings for patent researchers and agents.
DataEngUtils — a platform for data engineering utilities.
esmc.cpp — a C++ inference engine for protein language models (maintainer).
PepEmb — peptide embeddings for biological sequence understanding.
AdTestPro — platform for testing and optimizing ad campaigns (archived).

writing

I write about ML engineering, technology, and lessons learned from building production systems. See the blog for technical insights and project deep-dives.

contact

If you have a project in mind or want to explore ideas, reach out by email or book a call.