arXiv:2606.00001v1 [cs.SE] · Bengaluru, India · 2026

Building AI Systems That Survive Production:
A Portfolio

Shreyash Pandey

AI Systems Engineer, IBM Software Labs · shreyash.pandey.katni@gmail.com

Abstract

This portfolio documents work at the intersection of applied AI, backend engineering, and reliability: a 125M language model trained from scratch on a single consumer GPU, multilingual Indic models, agentic publishing and security systems, and a three-tier locator auto-healing system shipped inside IBM's test automation. We report honest benchmarks and provide four interactive artifacts that run entirely in the reader's browser.

1Selected results

The systems below were evaluated under internal or public benchmarks. Numbers are reported as measured, including where they are near chance.

MetricValue
test locators auto-healed5K+
availability supported99.99%
fewer support tickets70%
TechInterrupt Hackathon, India1st
Chrome fallback accuracy83%
tokens trained from scratch2B
Table 1. Headline metrics across shipped and research systems.

2Systems

2.1  Phoenix 125M (Released)

A LLaMA-style 125M decoder-only model trained from scratch on a single RTX 3080 Ti with a custom tokenizer, data pipeline, and training loop.

PropertyValue
Parameters125M
Tokens~2B
WinoGrande0.507
LicenseApache 2.0
Table 2.1. Phoenix 125M — PyTorch, RoPE, SwiGLU, RMSNorm, FlashAttention, BPE 32K. artifact ↗

2.2  Sweta-Hi / Sweta-Kn (Released)

Hindi and Kannada pretraining on a LLaMA-style architecture with a joint 64K tokenizer and an end-to-end multilingual data pipeline.

PropertyValue
Parameters134M
Vocab64K
Hindi PPL14.5
Kannada PPL34.0
Table 2.2. Sweta-Hi / Sweta-Kn — PyTorch, 64K BPE, Sangraha, Samanantar, IndicCorp v2. artifact ↗

2.3  LinkedIn Post Swarm (Live)

A multi-agent publishing workflow: a research agent, a Claude orchestrator, a critic loop, and a Playwright publisher with a Telegram human-in-the-loop.

PropertyValue
Engagement5x in 1mo
Follow reqs~1/day
Agents4
Table 2.3. LinkedIn Post Swarm — Claude, Qwen2.5 14B, Ollama, Playwright, FastAPI. artifact ↗

2.4  Rudra (Parked)

A multi-agent offensive-security architecture with hard-coded scope guardrails, sandboxed execution, and auditable event-driven workflows.

PropertyValue
Agent types4
Implemented2 / 4
Cluster VRAM22 GB
Table 2.4. Rudra — FastAPI, Ray, LangGraph, Kafka, Qdrant, Docker. artifact ↗

2.5  LocalLeads (Live)

End-to-end local-business discovery, AI site generation, deployment, and personalized outreach. A multi-hour agency process compressed into minutes.

PropertyValue
Site gen< 2 min
Cost/siteRs 500
Contacted~25
Table 2.5. LocalLeads — Claude API, Ollama, Playwright, Netlify, SQLite. artifact ↗

3Interactive artifacts

All four artifacts execute client-side. No inference server is contacted; computation runs in JavaScript on the reader's device.

· 0 tokens
try:

Figure 1. BPE tokenization under the Phoenix 32K English vocabulary versus the Sweta 64K joint vocabulary. Note the fragmentation of Indic scripts under the English-only tokenizer.

Figure 2. Semantic retrieval over the project corpus via TF-IDF and cosine similarity, computed in-browser.

app.under.test
Test step: click( locator: "#login-submit" )
Tier 1 · CSS selectorExact match on #login-submit
Tier 2 · Text embeddingNearest neighbour on label + nearby text
Tier 3 · Granite 3.3 VLMFind the primary action button visually

Figure 3. The three-tier locator auto-healing ladder. Recovery escalates from exact CSS match to text embedding to a vision-language model.

Q1 / 5 · You 0 · Phoenix 0

Figure 4. Commonsense reasoning game versus Phoenix-125M. The model scores ~0.507 on WinoGrande (chance = 0.50).

4Experience

Software Development Engineer 2, IBM Software Labs (2024 - Present). Reliability-heavy AI capabilities inside browser automation and testing products: embeddings, vision-language models, service decomposition.

Software Engineer 2, Software AG (now IBM) (2023 - 2024). Semantic retrieval, internal copilots, and prediction systems. Won the TechInterrupt Hackathon, first in India.

Software Engineer, Software AG (2022 - 2023). Enterprise integration platform work across Java, Spring Boot, and REST APIs. 2023 Star Performer.

References

  1. Shreyash Pandey. HuggingFace model releases. https://huggingface.co/shreyash-Pandey-Katni
  2. Shreyash Pandey. Source repositories. https://github.com/shreyash-Pandey-Katni