Sourish Sarmah
Sourish Sarmah

Senior Software Engineer

Sourish Sarmah

Based in India · Building at Agentic Private Markets

I'm a software engineer with six years of experience building production backends — most recently focused on real-time orchestration, LLM tooling, and high-throughput data infrastructure. I write primarily in Python and Go, with the occasional detour into C++ and Swift.

Today I lead backend engineering for agentic AI products at Agentic Private Markets, where I architect multi-agent research pipelines and the streaming inference engines that power them. Before that I led platform engineering at Tailorbird, shipped serverless healthcare infrastructure at Biofourmis, and built data pipelines at Inviz AI.

I care about systems that stay quietly correct under pressure — the kind that don't make it into release notes because they don't break.

Experience

  1. 2025 — Now

    Senior Software Engineer · Agentic Private Markets

    Building the agentic backbone of a private-markets research platform — orchestration engines, multi-agent pipelines, and streaming inference layers that turn unstructured filings into structured intelligence.

    • Engineered a multi-agent research orchestration pipeline that decomposes complex queries with LLMs, executes parallel web and file searches, and synthesizes responses within strict token budgets.
    • Architected a real-time AI agent orchestration engine processing 50+ streaming event types from OpenAI's Responses API, with dynamic tool resolution via MCP servers and Redis-backed conversation state.
    • Built a financial Artifact Agent that turns raw private-company data into auto-generated charts, tables, and comparisons via modular renderers and RAG-based file search.
    Python · FastAPI · LangGraph · OpenAI · MCP · Redis · PostgreSQL
  2. 2023 — 2025

    Senior Software Engineer · Tailorbird

    Owned the platform end-to-end for a construction-tech company — RFP portals, AI document parsing, and the plumbing that turns a fragmented contractor workflow into one cohesive product.

    • Spearheaded the RFP portal end-to-end, streamlining the bidding process for construction contractors and consolidating fragmented vendor workflows.
    • Designed Magic Folder, an AI document parser leveraging LlamaParse and RAG over OpenAI GPT models to extract critical insights from property assessment reports.
    • Shipped an LLM workflow that extracts line items from contractor invoices and auto-loads them into projects, reducing manual processing time by 90%.
    • Provided technical leadership for a team of three engineers, owning architecture decisions and the code-quality bar.
    Python · Django · PostgreSQL · LlamaParse · OpenAI
  3. 2022 — 2023

    Software Engineer · Biofourmis

    Worked on remote patient monitoring infrastructure where latency, correctness, and uptime were non-negotiable. Most of the work here was about extracting orders of magnitude from systems that already existed.

    • Built a serverless sync architecture for mission-critical medical data using AWS Lambda, API Gateway, and AWS SAM.
    • Reduced system latency 25× via Redis-based caching for hot paths.
    • Tuned PostgreSQL TimescaleDB schema and indexes — yielding 60–90× faster queries.
    • Migrated a microservice from Python to Go, resulting in 20× faster API responses.
    Go · Python · AWS Lambda · TimescaleDB · Redis
  4. 2020 — 2022

    Software Development Engineer · Inviz AI

    First role out of university — across data platforms, MLOps, and real-time services. The breadth was the gift; everything since has been a sharpening.

    • Built a Socket.IO microservice delivering real-time notifications by consuming queued messages and managing live socket sessions.
    • Shipped an async export feature for large-scale data to cloud storage in multiple formats using multipart upload.
    • Integrated a custom ML model training feature for Loopr through an Azure ML Studio MLOps pipeline.
    • Built data platform and ETL pipelines for TataCliq, powering several production ML models via AWS Glue.
    Python · Socket.IO · Azure ML · AWS Glue · PySpark

Selected Projects

Stack

Languages
Python, Go, C/C++, SQL, JavaScript
AI & LLM
LangChain, LangGraph, OpenAI Responses API, MCP, RAG, LlamaParse
Frameworks
FastAPI, Flask, Django, Node.js, PySpark
Backend Tools
Socket.IO, Kafka, RabbitMQ, Airflow, Docker, Git
Databases
PostgreSQL, TimescaleDB, MongoDB, DynamoDB, Redshift, MySQL, Redis
Cloud
AWS, Azure, GCP

Education

2016 — 2020

B.Tech, Computer Science & Engineering

Tezpur University, India · GPA 8.3 / 10.

Contact

Open to senior backend & AI engineering roles, technical leadership opportunities, and considered collaborations. Email is the fastest channel.