Skip to content
View aashi2912's full-sized avatar
9 followers · 10 following

Achievements

Achievement: QuickdrawAchievement: Pull Sharkx2

Achievements

Achievement: QuickdrawAchievement: Pull Sharkx2

Block or report aashi2912

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
aashi2912/README.md
Typing SVG

LinkedIn Portfolio Email

✦   AI Product Manager · SWE background ✦

🧠 The short version

I build enterprise AI products that turn complex research and data workflows into tools people actually want to use.

My edge: I was a software engineer first - 2 years building real-time analytics platforms and trading tools. That means I don't just write requirements and hope for the best. I understand what's feasible, where the technical risk lives, and how to have real conversations with the teams building the thing.

Now I work at the intersection of LLMs, RAG, agentic workflows, and enterprise product delivery - shipping secure, compliant B2B solutions with measurable outcomes.

📍 Right now

Role    → AI Product Manager
Domain  → Generative AI + Research Tech
         + Alternative Data
Stack   → LLMs · RAG · AI Agents · NLP
         Embeddings · Vector Search
Eng bg  → React · Next.js · Python · SQL
         Node.js · MongoDB
Focus   → Enterprise, secure, shippable

⚡ Impact - by the numbers

  What changed How
🚀 60% faster time-to-insight for end users LLM-powered automation of high-frequency research workflows
⏱️ 12 hours → under 2 hours on critical workflows AI agents handling data-intensive manual tasks end-to-end
📉 60% reduction in manual research effort Agentic RAG pipelines replacing repetitive analyst work
📈 25% efficiency gain across teams Research technology enhancements across cross-functional delivery
⚙️ 40% platform performance boost Optimised caching, API architecture, and data-fetching strategies
👥 8,000+ enterprise users served Multi-agent GenAI platform featured by NVIDIA

🛠️ What I work with

AI / ML

LLMs RAG AI Agents Agentic Workflows NLP Embeddings Vector Search Prompt Engineering Model Evaluation Responsible AI

Product

PRD OKRs 0-1 Roadmap GTM A/B Agile User Research

Technical

Python SQL JavaScript React Next.js Azure Node.js JIRA Figma

🧪 Things I've built

🗺️ Vibe Route   Live

Google Maps tells you the fastest route. This tells you the best one.

AI-powered walking route comparison that scores routes by how they feel — not just how fast they are. Built with LLM scoring, DBSCAN clustering, and NLP sentiment pipelines.

Claude API   DBSCAN ML   NLP   Product Discovery

📺 Blind Spot   Live

80% of what you watch was chosen by an algorithm. This shows you what it's hiding.

Surfaces filter bubbles in Netflix viewing behaviour using cosine similarity, content-based filtering, and the Herfindahl index to measure recommendation diversity.

Claude API   Cosine Similarity   Herfindahl Index   AI

💡 How I think about product


"The best AI products don't feel like AI. They feel like someone finally understood the problem."


Ship small, learn fast, kill what's not working — and have the courage to say so.

I believe I don't believe
🎯 The user's real problem > the stated requirement 🚫 AI is a solution looking for a problem
📐 A good PRD saves 10x the engineering time 🚫 Moving fast means skipping clarity
🔍 Talking to users is never optional 🚫 Internal proxies understand end users
🛡️ Security and compliance aren't afterthoughts 🚫 You can bolt guardrails on later
📊 If you can't measure it, you didn't ship it 🚫 Adoption is someone else's problem

I don't believe in building features nobody asked for. For me, AI isn't the product — it's the tool. The product is the problem you're solving, and how seamlessly you solve it. If the user has to think about the AI, you've already lost.

Always up for conversations about AI product strategy, research tech, or building things that actually work.

Let's Talk Connect Portfolio

Pinned Loading

  1. aashi-portfolio aashi-portfolio Public

    TypeScript

  2. aashi2912 aashi2912 Public

    Config files for my GitHub profile.

  3. blind-spot blind-spot Public

    A tool that shows you what Netflix's algorithm is hiding. Measures filter bubbles using cosine similarity, content-based filtering, and the Herfindahl index.

    JavaScript

  4. cn_tower_bot cn_tower_bot Public

    Python bot using Claude API + Twilio that sends a daily WhatsApp message at 8pm with a fun fact about that night's CN Tower lighting cause. ~100 lines, under 1¢/day.

    Python

  5. vibe-route vibe-route Public

    AI-powered walking route scorer — scores routes by how they feel, not just how fast. Uses Claude API, DBSCAN clustering, and 6 peer-reviewed studies.

    JavaScript