I build backend systems that work at scale. Previously worked with Python, Go, real time communication, infra, and production ready AI products.
I'm currently focused on:
- Building a scalable real-time leaderboard using SSE.
- Voice agent evals at work.
- Diving deep into databases, contributing to open source.
- Real time leaderboard with 28,232 concurrent SSE connections
- Optimizing Docker image builds for image size and build time
- Building a python package to turn unstructured data into financial insights
- Understanding Server Sent Events (SSE)
Recently released concall-parser, a python package to get structured features from earnings call reports. Made in collaboration with Jay Shah.
-
Send ranking updates to 28k+ concurrent SSE clients using Go, Redis Sorted Sets, and Prometheus. Optimized Docker images(48x size reduction, 46% build time reduced), structured logs, and system metrics for observability. Tech: Go, Redis, Prometheus, Docker, Grafana, PostgreSQL
-
Multi-Agent Recommender System
Built for a business-matching use case - uses OpenAI LLMs for DB query generation from natural language. FastAPI backend + MongoDB search with cron jobs to sync with Zapier tables. Tech: Python, FastAPI, MongoDB, Docker
-
Python package to extract structured data from earnings call transcripts using hybrid regex + LLM. 480+ downloads. Built with test automation (pytest regressions) & CI on GitHub Actions. Tech: Python, LLMs, Regex, GitHub Actions
- Languages: Python, Go
- Frameworks: FastAPI, Gin, Celery
- Infra & DevOps: Docker, AWS, GitHub Actions, Prometheus, Grafana
- Databases: MongoDB, Redis, PostgreSQL
- ML/Data: PyTorch, Keras, Weights & Biases, Scikit-learn, SpaCy, PyTorch Geometric, Tensorflow
- Data Pipelines: Apache Airflow, Apache Kafka
- Databases: Elasticsearch
I write about building AI tools, scaling real-time systems, and the innumerable sidequests I do.
I keep a list of blogs and discussions I found interesting/learned from at this post.