sagar
Internship Devlog

Founder × Intern

A high-fidelity chronicle documenting my 2-week AI/ML & Prompt Engineering residency at Virtual Height. Mapping the interface between non-deterministic generative intelligence and deterministic system integration.

Duration2 Weeks (June '26)
Role / FocusAI/ML & Prompt Engineering
Host CompanyVirtual Height Pvt Ltd

Featured Logs

Selected Highlights
Day 00Onboarding

Onboarding & Exploration: Mapping the Core AI & Python Blueprint

Onboarding at Virtual Height, introducing the training program, providing an overview of the curriculum, and mapping the 2-week schedule.

May 29, 20264 min readRead Log
W 01Reflections

Weekend 01 Reflection: The Corporate Loop & Majority Wins

Reflecting on the first five days of the internship. Realizing that while the technical teachings were profoundly basic, the true lesson was observing the endlessly 'cooked' corporate loop and the psychology of environments.

Jun 07, 20264 min readRead Log
FinalReflections

Epilogue: Final Reflections and Next Steps

A final retrospective on the 2-week residency. Comparing self-directed building with structured pacing, and closing the internship chapter to focus on next ventures.

Jun 15, 20264 min readRead Log

Internship Timeline

Days 01 - 14
Day 00 • May 29, 2026

Onboarding & Exploration: Mapping the Core AI & Python Blueprint

Onboarding at Virtual Height, introducing the training program, providing an overview of the curriculum, and mapping the 2-week schedule.

AIPythonPrompt-Engineeringonboardingsandbox-setup
Day 02 • Jun 02, 2026

Day 02: Python Control Flow, Functional Lab, and the Birthday Aura Hack

Setting up Python and VS Code envs, implementing conditional systems and match-case control flows, scripting a modular calculator, and presenting a surprise birthday web build to Rashmi.

Python-BasicsControl-FlowMatch-CaseFunctions-LabAura-FarmingBirthday-Surprise
Day 03 • Jun 03, 2026

Day 03: String Typology, Rule-Based Chatbots, and the Whiteboard Lecture

Analyzing list structures and dictionary lookups for chatbots, mastering string operations, and being selected by Rashmi to lecture the cohort on the whiteboard.

Python-CollectionsDictionariesChatbot-LogicString-NormalizationWhiteboard-SessionCohort-Gap
Day 04 • Jun 04, 2026

Day 04: APIs, JSON Schemas, and the Gemini AI Integration

Interfacing Flask with the Google Gemini API to build a local chatbot service, and mastering request-response structures and JSON payloads.

APIsJSON-SchemasFlaskGemini-APILocal-Server
Day 05 • Jun 05, 2026

Day 05: Prompt Engineering Mechanics, Hyperparameters, and the AGI Paradigm

Deconstructing LLM prompting structures (RTS), tokenization mechanisms, softmax temperature control, and exploring training, fine-tuning, embeddings, and RAG pipelines.

Prompt-EngineeringRAGEmbeddingsFine-TuningTokensHyperparameters
Day 06 • Jun 08, 2026

Day 06: HTML + CSS for Chatbot UI

Transitioning to frontend interfaces by mastering HTML structures, CSS styling for chat bubbles, responsive design, and demonstrating the SagarAI chatbot.

HTMLCSSChatbot-UIResponsive-DesignSagarAI
Day 07 • Jun 09, 2026

Day 07: Flask Fundamentals

Building backend bridges with Flask, handling HTTP requests, rendering HTML templates, and realizing the massive experience gap between traditional learning and production engineering.

FlaskBackendPythonHTTP-RequestsHugging-FaceDev-Gap
Day 08 • Jun 10, 2026

Day 08: The AI Chatbot Project & Architectural Pragmatism

Integrating remote LLM Inference APIs with Flask to build a custom StudyBot, analyzing pipeline overhead, and handling execution exceptions with robust error handlers.

HuggingFace-APIQwen-LLMFlask-IntegrationError-HandlingStudyBotArchitectural-Pragmatism
Day 09 • Jun 11, 2026

Day 09: Project Enhancement, State Management, and the Single-Page Chat Interface

Implementing session-based concurrent chat history storage in Flask, building a modular multi-personality system prompt compiler, and upgrading the user experience with asynchronous API requests.

State-ManagementFlask-SessionsAJAXFetch-APISystem-PromptsSingle-Page-App
Day 10 • Jun 12, 2026

Day 10: Final Project Presentation & Evaluation

Demonstrating the capstone StudyBot project before the Virtual Height evaluation jury, and defending system architecture choices during the final evaluation.

Final-ProjectSystem-DemoViva-Q&AFlask-Server

Ecosystem Takeaways

Key Learnings

  • Gaining clarity on the speed gap between self-directed learning and classroom pacing.
  • Realizing the funnel structure of introductory corporate training programs.
  • Constructing standard full-stack project directories (static assets, Jinja HTML templates, routing layers).
  • Implementing concurrency-safe session state management to isolate user conversation histories.
  • Designing request-response data flows between front-end inputs, back-end servers, and remote AI APIs.
  • Instantiating local Flask server environments and declaring URL routes.
  • Structuring webpage elements using HTML (forms, inputs, buttons).
  • Recognizing the deep exhaustion and repetitiveness of the standard corporate 9-to-5 loop.

Tools & Stack

  • Self-Reflection
  • Sociological Observation
  • Flask
  • Flask Sessions
  • HTML
  • Gemini API
  • Python 3.11
  • Internship Guidelines
  • Venture Mapping
  • Psychology
  • API Integration
  • Fetch API

Challenges Overcome

  • Balancing self-directed building with the slower pace of institutional training.
  • Navigating indirect promotional upselling pitches during final project evaluations.
  • Explaining complex stateless REST concepts clearly to beginners and senior evaluators alike.
  • Securing session cookies and ensuring they modify state properly on nested object updates.
  • Managing API connectivity limits and handling request connection timeout exceptions gracefully.
  • Debugging Method Not Allowed (405) errors when forgetting to explicitly allow POST methods on routes.

Join My Building Journey

I document and build in public, posting detailed AI architectural breakthroughs, product designs, and studio devlogs across my ecosystem channels. Connect on LinkedIn for collaborative insights.