Software & Data Engineer — Pune, India
I build high-performance, data-driven software — from C++ engines to end-to-end data pipelines.
scroll to walk the hall
About
Computer Engineering graduate (PICT '26, CGPA 8.8/10) who builds high-performance, data-driven software. My range runs from a C++ order matching engine to end-to-end ETL pipelines — bridging core systems engineering and the modern data stack.
Currently sharpening LLM evaluation and benchmarking as an AI Evaluator across leading platforms. Immediately available for Software Engineering, Data Engineering, or Data Analytics roles.
Experience
Scale AI · Outlier · Deccan AI
Evaluate LLM-generated code for correctness and complexity; benchmark models across DSA, agent reasoning, and CI environments.
Pune Institute of Computer Technology
Built a drag-and-drop workflow designer (React, TypeScript, React Flow) with a JSON serialization engine to a backend DSL — custom node/edge renderers, inline editing, and conditional routing.
Skills & Stack
C++ · Python · SQL · TypeScript · Java
Next.js · React · Flask · Streamlit
PostgreSQL · MongoDB · MariaDB · Snowflake
ETL Pipelines · Data Modeling · Apache Airflow · dbt · Batch Processing · Apache Spark (Basic)
DSA · OOP · Operating Systems · DBMS · Systems Design
Git/GitHub · Docker
LLM evaluation & benchmarking
Projects — the machines on this floor
Turning a rough idea into a shareable architecture diagram takes tools and time — this does it from a prompt, collaboratively, in real time.
Next.js · TypeScript · Gemini · Liveblocks · Prisma · Clerk
Two background AI pipelines (diagram generation + Markdown spec export) with real-time multi-user sync of nodes, cursors, and presence.
Exchanges must match orders fairly and fast under high-frequency load — this implements that core mechanism correctly and efficiently.
C++ · STL
O(log n) price-time priority matching via priority queues, partial fills with reinsertion, execution logs and real-time PnL.
Analytics teams need reliable, validated, queryable data — not raw transaction dumps.
Airflow · dbt · Snowflake · Python · SQL
Incremental loading in Snowflake, dbt data-quality validations, integration-level pipeline tests, monitoring with retries and SLA tracking.
Manual data cleaning is slow and unrepeatable — this makes it automated and auditable.
Python · OpenAI API · Pandas · SQL
LLM-generated cleaning rules applied as structured JSON plans with reproducible data-quality reports and normalized SQL schemas.
After hours
Contact
Immediately available for software, data engineering, and data analytics roles.
pranavdoke634@gmail.com