Bishal Rana
Technical Lead & Senior Python Developer — Courtroom Insight
About
Bishal Rana is a Technical Lead and Senior Python Developer with over five years of experience building scalable backends, AI-powered platforms and distributed data pipelines. He currently leads technical work at Courtroom Insight, where he designs legal-data pipelines, enterprise integrations and the microservices behind them.
His work sits where backend engineering meets applied machine learning: large-scale ingestion and ETL, entity resolution across messy real-world datasets, and search built on semantic embeddings rather than keyword matching. This site is his portfolio, built as a drivable Formula 1 circuit — each pit-stop on the track opens a chapter of the same career summarised below.
Experience
Technical Lead — Courtroom Insight
Building distributed legal-data pipelines, enterprise integrations and scalable microservices in Python on AWS.
- Reached 98.7% accuracy on automated legal-case matching using fine-tuned Sentence Transformers.
- Built ETL processing 88,000+ records per hour at under 0.1% error rate.
- Cut API latency from 12 seconds to 800ms — 93% faster — through query indexing, AWS RDS read replicas and ElastiCache.
- Ran fault-tolerant SQS queue workers on EC2 to keep ingestion resilient under load.
Selected work
Matter Connector Platform
An AI-powered legal intelligence platform at Courtroom Insight. It handles large-scale ingestion, normalization, mapping and search across millions of records drawn from many enterprise sources. Entity matching runs on BERT, Sentence Transformers and semantic embeddings, with ETL and entity-resolution workflows that detect and merge duplicate legal entities across distributed datasets.
Entity resolution is the hard part of a platform like this. The same firm, expert or matter arrives spelled a dozen different ways across a dozen different systems, and exact-string matching fails immediately. Embedding records into a shared vector space and comparing them semantically is what takes case matching to 98.7% accuracy, and it is what lets duplicates be merged automatically rather than by hand.
Beyond Exams
Part-time backend and data engineering for an EdTech learning platform, alongside the early computer-science work that started the career. Smaller scale than the legal-data work, but the same shape of problem: get data in reliably, model it sensibly, and keep the API fast enough that nobody notices it.
How Bishal works
Most of the engineering above comes down to three recurring problems, and the approach to each is deliberate.
Throughput. Moving 88,000 records an hour at under 0.1% error is less about raw speed than about failure handling. Fault-tolerant SQS queue workers running on EC2 mean a bad batch retries instead of taking the pipeline down, and the error budget stays where it should be.
Latency. The drop from a 12-second API response to 800ms — 93% faster — came from treating the database as the primary suspect: query indexing first, then AWS RDS read replicas to move read traffic off the writer, then ElastiCache for the results that did not need recomputing. No rewrite, just measurement in the right order.
Correctness at scale. Machine learning is only useful here if it is measurable, so accuracy is tracked as a number rather than a vibe, and the matching models are fine-tuned against real data instead of used off the shelf.
Skills
- Languages: Python, Java, SQL, PHP
- Backend: Flask, FastAPI, REST APIs, WebSockets, microservices
- Data: Pandas, NumPy, ETL pipelines, entity resolution
- Cloud: AWS — SQS, RDS, EC2, ElastiCache
- Machine learning: NLP, BERT, Sentence Transformers, embeddings, vector search
- Databases: MySQL and MariaDB, including performance tuning
- Engineering: pytest, CI/CD, Linux, Git
Frequently asked
Who is Bishal Rana?
Bishal Rana is a software engineer working as a Technical Lead and Senior Python Developer at Courtroom Insight, with more than five years of experience in backend engineering, AI-powered platforms and distributed data pipelines.
What does Bishal Rana work on?
Distributed legal-data pipelines, enterprise integrations and scalable microservices — most visibly the Matter Connector Platform, an AI legal intelligence product built on Python, FastAPI and AWS, with NLP entity matching over millions of records.
What technologies does Bishal Rana use?
Primarily Python — with Flask and FastAPI on the web layer, Pandas and NumPy for data work, MySQL and MariaDB for storage, and AWS (SQS, RDS, EC2, ElastiCache) underneath. On the machine-learning side: BERT, Sentence Transformers, semantic embeddings and vector search. He also works in Java, SQL and PHP.
How do you contact Bishal Rana?
By email at bishalranabr7@gmail.com, or through the GitHub and LinkedIn profiles listed below. He is happy to talk about building something or just to talk shop.