iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Building useful AI applications, in Unni T A’s account, means designing more than a prompt and a model response. His seven projects explore the surrounding system: tools, data access, retrieval, memory, evaluation, security, and application workflows. The projects range from research and financial analysis to appointment automation and enterprise retrieval, with one important distinction: some are described as functional applications, while another is explicitly an architecture demonstration with major components still missing.
This is a first-person portfolio overview, not an independent audit of the projects or their production behavior. It was published on DEV Community on September 29, 2026.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
What the projects explore
The projects are easiest to understand by the job they attempt to do. Some use an agent to plan and call tools; others center on retrieval-augmented generation (RAG), workflow automation, or a combination of structured database queries and document search. The descriptions below reflect the author’s account in his DEV Community article; they should not be read as independently verified implementation or performance claims.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →1. Deep Research Agent: gather, check, and organize information
The Deep Research Agent is described as a multi-stage research workflow. It accepts a topic, generates search queries, collects information, extracts facts, identifies gaps, and performs follow-up searches before producing a structured report. The author names FastAPI, LangGraph, Tavily, and ChromaDB in its stack.
#1 Best Overall
Beyond the initial report, the project is described as supporting exports, follow-up questions, summaries, counterarguments, translations, and job history. The author also lists Docker deployment and an Ollama-based local mode. These features make the project an example of an agent built around a research process rather than a single answer-generation call.
2. Autonomous Financial Research Agent: combine sources and analysis
This project is presented as a ReAct-style agent for financial research. Its described inputs include SEC EDGAR filings, earnings-call transcripts, financial data, news, and sentiment. The author says it can perform peer comparisons, fact-checking, and calculations.
The reported memory design separates working memory, semantic memory using FAISS, and episodic memory for prior runs. The author also describes conflict resolution, personally identifiable information (PII) redaction, prompt-injection protection, rate limiting, and evaluation. The article does not provide independently verified accuracy or financial-analysis results.
3. Autonomous Dental Appointment Bot: connect conversation to operations
The dental bot is described as handling appointment booking, rescheduling, and cancellation through web, SMS, WhatsApp, and voice. Its named components include PostgreSQL, Redis, Celery, Stripe, and Google Calendar.
Rather than focusing only on the conversational interface, the project description highlights operational concerns: slot locking, payment webhooks, duplicate calendar events, logging, health checks, and error handling. Those details illustrate the kind of application logic needed when an AI-facing feature can trigger changes to appointments and payments.
4. NexusBase: enterprise retrieval architecture
NexusBase is presented as an enterprise RAG architecture using Next.js, FastAPI, LangGraph, PostgreSQL, and pgvector. The author describes query routing and retrieval evaluation as part of the design.
At the time of the article’s publication, its backend was labeled functional and its frontend was being redeployed. Those are publication-time status descriptions, not confirmation of the project’s current state.
5. MedComply: document workflows for a compliance-oriented SaaS
MedComply is described as a medical-compliance SaaS monorepo with a Next.js frontend, FastAPI backend, and Supabase migrations. The listed application areas include organizations, users, documents, authentication, role-based access control, document processing, and AI-assisted analysis.
Rank #3
The description indicates a broader product structure around document analysis: user and organization management, access controls, and processing workflows are part of the project, not merely an AI interface. The article does not establish regulatory certification or compliance outcomes.
6. Aequitas FI: pair structured finance data with document retrieval
Aequitas FI is described as combining structured financial data and RAG. Its central design distinction is that SQL analysis is separated from document retrieval, with LangGraph, PostgreSQL, and pgvector named in the architecture.
The author also reports temporal comparisons, PII redaction, audit logging, human feedback, and automated testing. In this project, retrieval from documents and analysis of structured data are presented as separate capabilities that can work within one application.
Free tools Windows power users keep installed
One-click scans. No signup required.
7. Context Synthesizer: an architecture demonstration, not a complete system
Context Synthesizer demonstrates a possible enterprise retrieval workflow across Slack, Jira, Google Drive, and Notion. The author explicitly says it is not a complete deployed system: live connectors, the vector database, the embedding pipeline, the backend retrieval engine, authentication, and production LLM inference are not implemented.
That distinction matters. The project can illustrate an intended architecture, but its described status does not support treating it as a working integration across those services or as a production-ready RAG system.
How the project patterns differ
The seven projects are not presented as competing products or ranked alternatives. Their descriptions instead show several ways to organize an AI application around its task.
| Project | Primary task | Described system pattern | Publication-time status or qualification |
|---|---|---|---|
| Deep Research Agent | Research and report generation | Multi-stage, tool-using research workflow with retrieval | Features and deployment modes are reported by the author; not independently audited |
| Autonomous Financial Research Agent | Financial research and analysis | ReAct-style agent with multiple sources, memory, and calculations | Accuracy and production behavior are not independently established |
| Autonomous Dental Appointment Bot | Appointment and payment workflows | Conversational workflow automation linked to operational services | Implementation details are reported by the author |
| NexusBase | Enterprise information retrieval | RAG architecture with query routing and retrieval evaluation | Backend labeled functional; frontend being redeployed at publication |
| MedComply | Medical-compliance document workflows | SaaS application with document processing and AI-assisted analysis | Regulatory certification or compliance outcomes are not established |
| Aequitas FI | Financial analysis across records and documents | Structured SQL analysis separated from document retrieval | Features are reported by the author |
| Context Synthesizer | Cross-tool enterprise context retrieval | Architecture demonstration for a potential RAG workflow | Connectors, database, embedding pipeline, backend retrieval, authentication, and production inference are not implemented |
The author’s main lesson: an AI application is a system
Unni T A frames the progression from simple AI interfaces toward applications that have to make decisions about their environment and behavior. In his words, “My goal is not just to make an LLM generate an answer.” He describes the work as deciding what information a system can access, which tools it can use, what it should remember, how it retrieves information, and how results are checked.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
His reflections also emphasize the less visible work around failure handling, deciding when an agent should stop, protecting sensitive information, and deploying the application. These are the author’s lessons from building the projects, rather than independently established findings about all AI systems.
Best Value
That perspective connects the portfolio’s varied examples. The research agent is organized around gathering and refining evidence; the financial projects combine analysis with records and documents; the appointment bot connects an AI interaction to scheduling and payment operations; and the RAG examples focus on retrieving organizational knowledge. Across them, the model is only one part of the product.
What this portfolio does—and does not—establish
The article offers a useful map of the kinds of problems its author has chosen to explore, and names technologies and design concerns for each project. It does not provide a common benchmark, comparative test results, or independent verification of the implementations. The publication-time labels for NexusBase and Context Synthesizer should not be taken as current status guarantees.
For readers evaluating the work, the most informative distinction is between the stated goal of a project and the implementation status the author actually reports. Context Synthesizer is expressly a demonstration with core production components absent; NexusBase is described with a functional backend and a frontend redeployment in progress at publication. The remaining descriptions outline reported functionality and architecture, not external proof of reliability, security, or real-world outcomes.
Recommended Free Tools
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
