You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Add First American experience, harnesskit, Helix, Jira extension to about page
Add full First American career (2017-2021, three roles).
Remove PrepForge from side projects.
Add harnesskit (agent framework), Helix (knowledge system), Jira Ticket Creator (VS Code extension).
Copy file name to clipboardExpand all lines: src/pages/about.md
+53-3Lines changed: 53 additions & 3 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -169,6 +169,44 @@ By the end of 2022, I was handling production deployments, doing code reviews, r
169
169
170
170
---
171
171
172
+
## 2017 – 2021
173
+
174
+
**First American (India)** · Angular, ASP.NET Core, .NET Core, SQL Server, Google Maps API, AWS
175
+
176
+
Four and a half years across three roles, all in Bengaluru. Started as a technical trainee, left as a software engineer building customer-facing products.
177
+
178
+
**Software Engineer** · Apr 2019 – Aug 2021
179
+
180
+
Properties Team:
181
+
- Piloted the project — set up the structure, created quick POCs that helped shape the product
182
+
- Built property search via map click, address search, and APN search
183
+
- Integrated xUnit framework for unit testing across projects
184
+
- Developed GlobalNav — a JavaScript library consumed at runtime across applications
185
+
- Designed reports: Property Detail, School Information, Transaction History using ActiveReports
186
+
187
+
Eagle-Status Team:
188
+
- Extended multi-select dropdown, created custom drag-and-drop directives using DOM APIs
189
+
- Optimized AgGrid to load large row counts at startup
190
+
- Set up HMR (Hot Module Reloading) in Angular with .NET Core
191
+
- Fixed 20+ critical UI bugs that unblocked initial release
192
+
193
+
**Member Technical Staff** · Aug 2017 – Mar 2019
194
+
195
+
MyFirstAm Team:
196
+
- Built zip code property search, REST APIs for property data and reports from DataTree
197
+
- Built an adapter layer for WCF to DTAPI migration
198
+
- Saved ~$80K/year by implementing Static Street View instead of Panoramic View for all calls
199
+
- Wrote Terraform scripts for AWS infrastructure as part of cloud migration
200
+
201
+
**Technical Trainee** · Feb 2017 – Aug 2017
202
+
203
+
Budget-Aid Team:
204
+
- Redesigned the UI using razor views and jQuery
205
+
- Introduced unit testing, increased test coverage by 20%
206
+
- Migrated the project from MVC to WebAPI
207
+
208
+
---
209
+
172
210
## What I build at night
173
211
174
212
### Jishu — a personal AI agent · Python, FastAPI, Claude API, SearXNG
@@ -213,11 +251,23 @@ Mostly did it to see how far .NET has come. It's come far.
An open-source framework extracting production agent patterns into reusable modules. Tool registry with Zod schemas and permission levels. Query loop (input → context → model call → tool calls → repeat). Permission engine with deny/ask/allow cascade. Context manager with cache-aware compaction. Agent spawner with worktree isolation. Hook system with lifecycle events. Model router with tiered selection and fallback.
257
+
258
+
The core idea: three intelligence levels — rule-based (no model, $0), hybrid (local model), and full agent (cloud model) — all running on the same architecture. The model is optional. The harness isn't.
259
+
260
+
---
261
+
262
+
### Helix — personal knowledge system · Markdown, Git, Python
263
+
264
+
The memory layer behind Jishu. A git repo of markdown files organized into four layers: immutable raw sources (work context, memories, plans, investigations), active state (goals, decisions, journal, feedback), personal context (rhythms, aspirations, relationships), and cognitive architecture (how Jishu thinks). Replaced a vector database (Milvus + Ollama embeddings) — prompt caching made flat file reads cheaper and more reliable than embed-retrieve-inject. [Wrote about the full design](/posts/giving-an-ai-agent-memory-that-survives-the-session).
265
+
266
+
---
217
267
218
-
An interview preparation system I'm building for myself. Three thousand DSA problems sourced from LeetCode with company-wise tagging, twenty-six system design scenarios, twenty object-oriented design problems. Uses FSRS (Free Spaced Repetition Scheduler) to surface problems based on when I'm most likely to forget them. AI-powered coaching for hints and explanations, mock interview simulation.
268
+
### Jira Ticket Creator · JavaScript, VS Code Extension API, Copilot LM API
219
269
220
-
Still a work in progress — building it because I couldn't find a prep tool that combined spaced repetition with the specific problem sets I wanted.
270
+
A zero-dependency VS Code extension for creating Jira tickets with AI assistance. Describe what you need, the extension generates summary, description, type, priority, and labels. Single and bulk ticket creation. Supports cookie auth for enterprise Jira behind SSO/IAP — built specifically to work with Motorola's Google Cloud IAP setup. Distributed to the team via GitHub Releases.
0 commit comments