The Road to Building AI Systems.
The Story So Far
-
Starting Out
Learning Through ProjectsI became interested in programming by exploring how software could handle repetitive tasks. That curiosity continues to influence the problems I choose to work on and the solutions I try to build.
-
Exploring AI
Learning Through ExperimentsI explored machine-learning ideas and large-language-model tools, then began applying what I learned through projects. I'm interested in understanding where these tools help and where their limitations matter.
-
Building Automation Projects
From Problem to ProjectInternPilot AI explores the process of collecting internship listings, removing duplicates, organizing opportunities, and presenting results in a digest. The focus here is the workflow, not an unverified claim about time saved or continuous operation.
-
Voice AI Experiments
PVCA · Workflow ProjectPVCA explores a missed-call workflow using Twilio, a voice agent, webhook processing, and Telegram summaries. It is presented here as a project, without unverified claims about response time, operating cost, or production availability.
-
Current Focus
B.Tech · Computer Science at SGTUI'm pursuing a B.Tech in Computer Science at SGTU. My public projects include machine-learning notebooks, data-analysis work, web applications, and experiments with AI and automation. The project pages link to public repositories and README documentation where available.
-
Future Vision
Agents That Reason, Not Just ExecuteThe next chapter is more sophisticated — multi-step reasoning, persistent memory, tool use, real-world integrations that handle edge cases gracefully. I want to build systems that don't just automate tasks but make intelligent decisions in context. The goal isn't automation for its own sake. It's giving people back hours they're spending on things that shouldn't require a human.
Milestones
-
Microsoft Elevate AI Program
2025Selected for Microsoft's Elevate AI program in 2025.
-
InternPilot AI — Project
Internship-discovery workflowAn internship-discovery project focused on collecting opportunities, removing duplicates, organizing results, and delivering a digest through Telegram.
-
PVCA 1.0 — Project
Voice-assistant workflowA voice-assistant project exploring missed-call handling, voice-agent interaction, webhook processing, and Telegram summaries.
-
Data Analyst @Unified Mentor
May – Jul 2025 · InternshipInternship experience in data analysis using Python and Power BI.
-
Frontend Developer @Edunet Foundation
Aug – Oct 2025 · InternshipInternship experience focused on frontend development and responsive interfaces using HTML, CSS, and JavaScript.
Built in Public
Live numbers from my public GitHub (@Soham-o) — refreshed automatically, never typed in by hand.
Languages across projects
- Jupyter Notebook50.7%
- Python38.4%
- JavaScript5.8%
- TypeScript2.3%
- CSS1.7%
- HTML1.1%
Contribution activity
287 contributions in the past year · longest streak 7 days
Recent development
- Pushed updates to EarthShare
- Pushed updates to AirTwin-X
- Pushed updates to Cursor Bite
- Pushed updates to PersonaLearn
- Created Cursor Bite
- Created AirTwin-X
Growth since tracking began (11 Oct 2026)
Verified on GitHub
Generated from public GitHub data by fixed rules (first launches, releases, star & fork milestones, streaks, merged contributions). Each item links to its source.
-
Open Source Contributions — 7 Merged PRs
Latest Oct 2026Pull requests merged into 3 public projects owned by others, including RogueAlg0/taken, hermes-labs-ai/lintlang, 1jehuang/jcode.
Verify on GitHub →
-
10 Public Projects — Milestone
Jun 2026Reached 10 public projects with Autonomous Driving System.
Verify on GitHub →
-
7-Day Streak — Consistency
29 Jan 2026 – 4 Feb 2026Contributed on GitHub 7 days in a row — the longest run in the past year.
Verify on GitHub →
-
First Public Project — Employee-Analysis
Jun 2025 · LaunchEmployee-Analysis became the earliest repository in this portfolio, created 23 Jun 2025.
Verify on GitHub →
GitHub data synced 25 Oct 2026
What Guides the Work
THINGS I ENJOY
Exploring new AI tools before they're mainstream. Turning "what if I automated this?" into a weekend project. Reading about agentic AI and multi-model reasoning. Finding the laziest — most elegant — solution to any problem. Explaining AI to people who assumed it was magic.
I have a running list of "repetitive things that should be automated." It is always longer than my project backlog. Always.
HOW I WORK
Prototype fast, then engineer properly. README-first: if I can't explain it, I don't fully understand it. Async and non-blocking wherever possible. Late-night debugging is often the most productive time.
My benchmark for a good system: if it breaks silently at 3 AM and nobody notices until morning, it was working correctly.
"I'm still early in this journey — but every system I ship is another proof that intelligent automation doesn't need a budget, a team, or a deadline. It needs a problem worth solving and the stubbornness to see it through."