Dhevenddra K G
Dhevenddra K G is an individual software engineer and solo developer/creator, not an organization. He builds enterprise platforms and applied-AI systems that are reliable, inspectable, grounded, and auditable, including code-intelligence, hybrid-RAG, and voice-agent projects.
At a Glance
- AI coding-agent and developer-tool users
- Enterprise software and platform engineering teams
- Healthcare technology and hospital software teams
- Service-desk and internal support operations
- +2 more
AI Tools by Dhevenddra K G
(1)
forensic-deepdive
Codebase Knowledge Graph for AI Agents
Discussions
No discussions yet
Be the first to start a discussion about Dhevenddra K G
Latest News
Published “Why I Switched from JSP to React (And What I Learned)” about migrating 50+ JSP pages in healthcare software.
Published “Building a Redis Cache Layer for Session Failover” describing distributed caching for JBoss EAP cluster nodes.
Published “What Working in Healthcare Tech Taught Me About Software,” reflecting on 2.5 years at Dedalus.
Released forensic-deepdive 0.10.0 on PyPI, described as “The Upgrade Path” integrity release.
Products & Services
Apache-2.0 multi-language static-analysis and software-forensics engine that creates a persistent code knowledge graph, models Git history, exposes nine composite MCP tools, and emits durable markdown artifacts and agent/editor shims. It supports nine languages and confidence-tagged EXTRACTED, INFERRED, and AMBIGUOUS relationships.
Provider-agnostic, local-first Tanglish voice service-desk agent with BM25 retrieval, working and long-term memory, guardrails, MCP-shaped tools, and a draft-confirm-execute workflow for state-changing actions.
Auditable insurance-claims adjudicator for the AMD Developer Hackathon: Act II. It returns APPROVE, FLAG, or DENY decisions grounded in verbatim policy clauses, with programmatically verified citations and SHA-256-backed audit records.
Streaming voice-activity detector and ASR front end built from hand-written DSP, adaptive gating, smoothing, a logistic-regression alternative, and pluggable Whisper/Saaras transcription adapters.
Market Position
The clearest documented competitive positioning is for forensic-deepdive: its README compares it with GitNexus, CodeGraphContext, DeepWiki, and Sourcegraph, emphasizing Apache-2.0 licensing, a persistent graph, confidence-tagged edges, Git archaeology, durable markdown artifacts, an agent-insight layer, multi-platform shims, and local-only operation. The portfolio overall differentiates its projects through grounded evidence, explicit uncertainty, local-first execution, and approval-gated actions rather than unconstrained generative behavior.
Founding Story
Dhevenddra's portfolio describes a personal focus on building systems that remain reliable and inspectable under real-world pressure. His independent work extends his professional Java/Python engineering into code intelligence, retrieval, voice systems, and local-first AI, with explicit guardrails and evidence rather than opaque model behavior.
Target Markets
- AI coding-agent and developer-tool users
- Enterprise software and platform engineering teams
- Healthcare technology and hospital software teams
- Service-desk and internal support operations
- Insurance claims and financial-services compliance teams
- Users needing local-first or offline-capable AI and communications
- Helping AI coding agents understand and navigate unfamiliar multi-language codebases
- Software architecture discovery, impact analysis, onboarding, refactoring, and Git-history investigation
- Grounded retrieval over proprietary enterprise frameworks and internal documentation
- Service-desk voice automation for Tamil-English/Tanglish users where actions require confirmation
- Insurance claims adjudication and financial-services compliance where evidence and auditability are required
- ESG obligation extraction and deadline tracking from loan agreements