Software Engineer - II (Language Engineering) at DeepSource (2025-08 – Present)
- Resolved core async and event-loop issues across Autofix and agent workflows, enabling stable parallel execution and improving reliability of automated code review pipelines.
- Built memory-aware AI review by retrieving repository memories from diff summaries and injecting ranked enforcement, suppression, and contextual signals into agent prompts.
- Implemented merge-aware incremental diffing using synthetic bases and hunk-level filtering, eliminating noise from base-branch merges and avoiding redundant re-analysis.
- Improved AI review precision through enhanced false-positive filtering, impact analysis, diff-context validation, accurate line attribution, and higher quality review explanations.
- Architected a staged AI code-review pipeline with parallel per-file review, selective Autofix orchestration, and bounded concurrency, balancing latency, token cost, and review quality through context-aware prompting.
Compiler Engineer at EdgeCortix (2024-06 – 2025-08)
- Developed optimized C99 TensorFlow kernels for an ARM-based MCU, achieving a 4% average speedup and up to 16% max speedup for Pack workloads over the original TFLite implementation.
- Designed a backend compiler-pass-based framework for validating intermediate model outputs, reducing debugging time by orders of magnitude.
- Built an automated x86-to-MCU testing pipeline for model validation, replacing manual test generation and accelerating MCU testing throughput by nearly 5X.
- Implemented multiple graph rewrite passes to map subgraphs onto target hardware accelerators, improving execution efficiency and reducing CPU fallback.
Associate Analyst at Hansa Cequity (2019-06 – 2020-07)
- Performed data wrangling in SQL Server using modular and optimized query building techniques, for further data analysis and sending personalized campaigns.
