GovSolve

GovSolve is a strictly asynchronous complaint redressal portal. When a local submits a grievance—in whatever way makes sense to them—the system doesn't try to chat. It acts as an intelligent processing engine. It bridges the gap between frustrated citizens and rigid bureaucratic workflows, ensuring every submission is understood, categorized, and made immediately actionable.

Role

Lead Engineer & Project Manager

Responsibilities

React.js,TypeScript,TailwindCSS,Python,Deep Learning
GovSolve Hero View

Architecture &
The Challenge

Handling government-level complaints means zero tolerance for AI hallucinations. The challenge wasn't generating responses; it was building a deterministic guardrail system around a non-deterministic LLM.

To handle the unpredictability of human input, I engineered a bifurcated encoder-decoder architecture. The Encoder acts as a ruthless classifier: it ingests the raw, unpolished submission, cuts through the noise, and maps it to the exact municipal department without stripping away critical context. Once classified, the Decoder takes over. But instead of generating generic AI fluff, the Decoder is tasked with two strict outputs: formulating a legitimate, measurable response for the citizen, and extracting structured, actionable insights for the government officials. It preserves the formal, authoritative essence of Philippine public service—no conversational padding, just rigorous data routing and clear resolution.

Visual Assets

Fig. 01 // GovSolve+
GovSolve artifact 1
Horizontal Asset 1
Fig. 02 // GovSolve+
GovSolve artifact 2
Horizontal Asset 2
Fig. 03 // GovSolve+
GovSolve artifact 3
Horizontal Asset 3

Outcome

"Transformed highly unstructured, emotional citizen inputs into clean, actionable data for municipal departments. It eliminated manual triage bottlenecks, provided leadership with measurable insights, and delivered formal, culturally accurate redressal to the locals."