Engineering Lead | Tech Startup Consulting | Xoogler

substack

github

medium

Akul M • Peerlist

<h3>LoopKit: A Finishing Harness for Claude Code</h3>LoopKit is a sophisticated finishing harness designed to enhance the reliability and control of AI code generation, specifically for Claude Code. It focuses on externalizing critical decision-making processes from the AI model into robust file, script, and hook mechanisms, preventing drift and ensuring consistent execution. This approach does not aim to make the AI smarter but rather to make its outputs more dependable by managing decisions that are prone to error under pressure.<ul><li>State-File Memory: Manages memory through state files, with state/triage.md acting as a queue and specs/ defining the truth. Findings are persisted to disk rather than being lost in conversation history.</li><li>Stage as a Lookup: The next stage of execution is determined by looking up the current state, ensuring a structured and predictable workflow.</li><li>Blocking Hooks: Implements critical hooks that prevent dangerous operations such as arbitrary file deletion, history rewriting, or premature merging of pull requests. These hooks enforce governance rules and ensure code integrity.</li><li>Stop Gate with Testing: A comprehensive stop gate mechanism runs prechecks, regression diffs, linting, type checking, and builds. It includes an acceptance review against specifications, with any failures blocking the process.</li><li>Separated Powers: Differentiates between implementer and reviewer agents, with the reviewer adopting a default stance of skepticism until proven otherwise. Merge and approval actions are restricted to prevent self-approval or unauthorized changes.</li><li>Escalation Door: Handles tasks requiring human intervention by routing them to an inbox (inbox/needs-human.md) with a structured format for review and decision-making.</li><li>Memory Management: Utilizes adapters for graph, memory, and knowledge retrieval, allowing the system to recall past decisions and context without relying on traditional Retrieval-Augmented Generation (RAG) methods.</li><li>Extensibility: Designed to be extended with project-specific skills, tools, and agents, allowing developers to integrate custom logic and workflows seamlessly.</li><li>Robust Testing: Includes comprehensive testing for the plugin itself, covering hooks, scripts, and end-to-end scenarios in a scratch repository to ensure reliability and correctness.</li></ul>LoopKit provides a framework for building more reliable and controllable AI-driven development workflows by enforcing strict processes and externalizing critical decision points.

LoopKit: a finishing harness

An ai powered search GPT for searching and displaying results with caching and synchronization capabilities.Hi everyone! I'm kicking off <a target="_blank" rel="noreferrer noopener" class="break-all link-break text-gray-500 hover:underline anchor_edefb8 anchorUnderlineOnHover_edefb8" href="https://github.com/salestracker/shodhan">a fun opensource project on how we search</a>. It is hacked together as a wrapper to build applications ontop of search. Think jarvis meets search. Looking for frontend devs, UI/UX designers to contribute. If you have done projects before, this project would give you an easy ramp up with a rich README to get set from day 0. If you are such a contributor, I am looking to connect and learn from you where would you like to take this forward? Feel free to reach out if you’re interested or have any questions! Key Features<ul><li>AI-Powered Search: Uses LLM API for comprehensive, cited answers</li><li>Conversational Interface: Supports follow-up questions with context preservation</li><li>Smart Caching: Redundant query prevention with localStorage-based caching</li><li>Modern UI: Built with Shadcn UI and Tailwind CSS</li></ul>Tech Stack: • Frontend: React, Vite, Tailwind, Shadcn #WebDev #ReactJS #OpenSource

Shodhan

Building Energy Management Consulting

Secure Judiciary Tools

Healthcare Commons for Digital Public InfrastructureRoughly 6 billion people or approximately 85% of the world's population <a target="_blank" rel="noopener noreferrer nofollow" class="text-green-text rich-link break-all link-break" href="https://www.ashmoregroup.com/sites/default/files/article-docs/MC_10%20May18_2.pdf">lives in an Emerging Market</a> country that lack streamlined reservation and appointment protocols for rapid offline healthcare consultation and outpatient services. Existing outpatient consultation has gaps between healthcare facility search and wait times in the queue during registration. This bottleneck is a contributing factor to critical capacity issues managing inpatient and outpatient flow.The difficulty lies in obtaining a clinical assessment or scheduling a medical appointment through existing reservation systems that lack hyperlocal triage.Our mission is to:<ol><li>Provide reduced wait times for consultation and out-patient services for at-risk patients.</li><li>Fix disproportionate allocation of resource constraints and inconsistent flow of patients.</li><li>Reduce wait times for dissatisfied patients</li><li>Reduce high healthcare administrative costs as a direct consequence of patient allocation and wait times.</li></ol>Project Healthcare, from hereon called the brokering service, is a set of upstream OpenAPI specification to extend any patient facing user interface looking to integrate to location-aware consultation and diagnostics facilities near them by:Exposing a diagnosis API to capture patient symptoms. Exposing geolocation API to caputure patient's location and offer nearby facilities to take action on based on their diagnosis and location.The specification follows a sandwich service model i.e. it requires one or more upstream producer services and one downstream consumer service as follows:The patient facing application, known from hereon as the downstream consumer service, calls the diagnosis and geolocation API. The brokering service stores transient diagnosis request and enqueues them to upstream provider service that should return records of facilities and their specialties based on the diagnosis. The brokering service returns the records of matching facilities to the downstream consumer service.Read More: <a target="_blank" rel="noopener noreferrer nofollow" class="text-green-text rich-link break-all link-break" href="https://github.com/Xcov19/project-healthcare/blob/main/README.md">README</a>

XCoV19

Using RF tag based simple UART driven communication, a prototype to demonstrate animal tracking in a forest area. Further thesis on extending communication to a solid state RADAR driven Animal classification and developing dsp algorithms on

Anti Poaching System Using Telemetry

(Under development)

The objective is to scan a definite given area & classify targets as slow moving, fast moving, clutter on the basis of a RADAR scan. Solid state RADAR based scanning systems are studied & appropriate system is interface