Backend & AIIndiaOpen to new roles

Sourish Sarmah

I build the backend systems that AI agents run on.

I'm Sourish Sarmah, a backend and AI engineer in India. I've spent six years building production systems, mostly in Python and Go, and the last few on LLM pipelines and real-time orchestration.

Until May 2026 I was Tech Lead at Agentic Private Markets, where I built their multi-agent research pipeline and a data grid that LLM agents query with SQL, and led the company's SOC 2 compliance work. Before that I owned the platform at Tailorbird. Earlier still: patient-monitoring infrastructure at Biofourmis and data pipelines at Inviz AI.

I like systems that stay correct under load. Nobody writes release notes about them, because they don't break.

Get in touch → Download résumé

Work

Six years of production backends.

2025–2026

Agentic Private Markets Tech Lead

Senior Software Engineer, Sep 2025 – Jan 2026 · Tech Lead, Jan – May 2026

Backend for a private-markets research platform. As tech lead I designed an Excel-like data grid with streaming ingestion, a SQL layer the agents call as a tool, and row-level security in the database, so LLM agents could run joins and aggregations over enterprise datasets without seeing another tenant's rows. I also led the SOC 2 work.

Before that I built the research pipeline: an LLM splits a question into sub-queries, runs web and file searches in parallel, drops duplicates and writes an answer inside a fixed token budget. I also shipped a voice agent on OpenAI Realtime that answers in under a second.

Python, FastAPI, PostgreSQL, LangGraph, OpenAI Realtime

2023–2025

Tailorbird Senior Software Engineer

Owned the platform at a construction-tech company and led a team of three. I built the RFP portal contractors bid through and a parser that pulls findings out of property assessment reports with LlamaParse and RAG. An LLM pipeline for contractor invoices cut manual processing time by 90%. I also added audit logging on database triggers, so every change is recorded.

Python, Django, PostgreSQL, LlamaParse, OpenAI

2022–2023

Biofourmis Software Engineer

Backend for a remote patient-monitoring product. I built serverless sync for medical data on AWS Lambda. Redis caching cut latency 25×, index and schema work made TimescaleDB queries 60–90× faster, and rewriting one service from Python to Go made its API 20× faster.

Go, Python, AWS Lambda, TimescaleDB, Redis

2020–2022

Inviz AI Software Development Engineer

My first job out of university. I built ETL pipelines and ML training datasets for TataCliq on AWS Glue, and a model-training pipeline for Loopr on Azure ML Studio.

Python, PySpark, AWS Glue, Azure ML

Projects

Things built on the side.

2026 · Video

Supaintro

Turns a product URL into a branded launch film: multimodal content generation, voice synthesis and Remotion rendering on autoscaling AWS Fargate workers fed by PostgreSQL job queues.

2025 · macOS

OCDify

Sorts and cleans up your Mac desktop with a quantized LLM running on the device. Your files never leave the machine.

Web · AI

Outstation

An AI travel planner. Give it a destination and it builds a day-by-day itinerary on a map that you can keep editing.

Linux

God's Eye

A Linux experiment that replaces the sudo password prompt with face recognition, hooked into PAM.

Stack

Mostly Python and Go, plus SQL. FastAPI, Django, Flask and PySpark; LangChain, LangGraph, the OpenAI APIs and RAG for agent work. Kafka, RabbitMQ and Airflow for moving data; PostgreSQL, Redis, MongoDB, DynamoDB and Redshift for keeping it. Shipped with Docker on AWS, Azure and GCP.

Education

B.Tech in Computer Science and Engineering, Tezpur University, 2020.

Contact

I'm looking for a senior backend or AI role. The fastest way to reach me is email.

LinkedIn GitHub Résumé