Prajwal HD

Software engineer focused on backend systems and applied AI. Three years building production infrastructure at Opslyft. I care more about how a system behaves under pressure than what it's built with.

Resume

Now

Updated April 2026

Looking for backend, platform, and applied AI roles with end-to-end ownership. California, remote, or hybrid.
Building QuackCode, a Linux-first desktop agent, and Gitwise, a CLI that writes commit messages from staged diffs.
Stress-testing agent workflows on real codebases: tool use, git safety, isolation, and recovery when the model gets it wrong.
Reading: Designing Data-Intensive Applications (re-read), papers on LLM tool use evaluation.

Featured work

Peer-reviewedMDPI Electronics · 2026

Re-Evaluating Android Malware Detection: Tabular Features, Vision Models, and Ensembles

Hosahalli Dayananda, P.; Chen, Z.

Compares tabular static features, vision-based byte-plot models, and ensemble strategies for Android malware classification. Evaluates where modality choice changes the detection surface and where ensembles actually pay for themselves.

Read paper

Projects

QuackCode

Electron desktop app for running coding agents against real repositories. Supports Claude through the local Claude CLI, Gemini through API keys, and Codex through codex app-server; streams tool activity live, persists threads and event history in SQLite, and can run in disposable git worktrees or scheduled automation threads.

Developer ToolsElectronGitSQLite

Gitwise

CLIGitTypeScript

Git helper that reads staged diffs, proposes repo-style commit messages, and turns plain English into branch names. Interactive by default, validates model output before showing it, and blocks outbound prompts when the diff looks like it contains secrets.

Multimodal Malware Detection

PythonMLSecurity

Master's thesis on whether combining static PE features with visual byte-plot images beats either modality alone on malware detection. Three-track evaluation over EMBER 2024 with matched splits and fusion ablations.

Guardscribe

PythonASRNLP

Streaming speech-to-text pipeline that transcribes live audio and masks toxic spans on the fly. Two-stage classifier keeps cost low on clean input, a token-level tagger redacts only the offending words, and the fast path stays under 400ms end-to-end on CPU.

All projects