ArjunVerma

I am a curious Engineer at heart. Been at this since I was twelve.

Find me elsewhere

GitHub↗

About

I build software I'd want to use.
Systems, interfaces, and the occasional weekend experiment.
I write to think clearly, and I ship to learn faster.

I use Arch btw.

02Selected work^[13]

You describe a concept in plain English and Stitcher writes the Manim code, runs it, and renders an actual animated video. Around 8K signups and 50+ people using it daily.
A crew of LLM agents handles codegen across 10+ providers and debugs its own mistakes in a retry loop. RAG over the Manim docs killed about 90% of the hallucinated code.
Every AI written script runs inside a locked down sandbox (bubblewrap + seccomp) so untrusted generated code can never touch the host machine.
Built a real time collaborative timeline editor with Yjs CRDTs over WebSockets. Basically Google Docs, but for editing video.
Payments are race safe: the Razorpay verify call and the webhook can both fire and you still get charged exactly once, thanks to an atomic Mongo update.

Attachments

Demo video↗

try it live!↗

TypeScriptPythonManimLangGraphLLaMa 3.1 8BRAGBullMQRedisMongoDBFFmpegRazorpay

Trained a 99M parameter transformer from scratch, same architecture family as Llama and Mistral (RoPE, grouped query attention, SwiGLU, RMSNorm).
Wrote my own byte level BPE tokenizer with a 32K vocab and built the full 4 billion token training corpus myself, deduping millions of near duplicate docs with a Bloom filter.
Full post training stack: supervised fine tuning, then DPO, with an LLM as a judge harness scoring relevance, persona and coherence on a held out set.
The whole thing runs client side in the browser (PyTorch to ONNX to WebAssembly, int8 quantized) at 100+ tokens/sec with nothing running on a server.
Taught it to use tools mid generation: it emits plain text calc tags the runtime catches and solves, so it does arithmetic cleanly and gets ~56% on harder word problems.

PyTorchTPUBPERoPEGQASwiGLUSFTDPOONNXtransformers.jsWASM

Built a 14 service agent system from scratch (gateway, main agent, 5 connectors, cron, a human in the loop dashboard) that runs 24/7 on a single machine.
Read the OpenClaw architecture and implemented its gateway myself: FIFO ticket dispatch over Redis, heartbeat based eviction, and recovery of in flight work if a worker dies mid task.
The core agent loads skills on demand, remembers past context with a vector memory (mem0), and can switch between LLM providers at runtime with automatic fallback when one fails.
Gave it real hands on the web with Skyvern for browser automation and Tavily for live research, so it can actually go do tasks instead of just talking about them.
Backed by a 381 test monorepo with real infrastructure integration tests spun up through Testcontainers, not mocks.

TypeScriptNode.jsBullMQRedisPostgresNext.jsPI Agentmem0SkyvernTavilyPM2

03Writing

10.07.2026·6 min read

Experience

• Built a multimodal ingestion layer (PDFs, DOCX, images, audio) with Docling + VLMs at 94% field-extraction accuracy. • Shipped 3-stage context compression keeping working memory under 8K tokens, cutting LLM spend 54%. • Built the FastAPI backend (35+ endpoints) and LangGraph pipeline turning pharma SOPs into lessons in 4 languages. • Sped up PDF parsing 150x (PyMuPDF) and lesson generation 7x (asyncio.gather).

• Rebuilt production RAG into a 5-stage pipeline (HyDE, rerank), lifting eval quality 4.12 to 8.93 and latency 12s to under 5s. • Built an LLM-as-judge harness and ran 33 experiments across 12 rerankers, 3 embedding models, and 5 architectures. • Curated and cleaned an 85K tutorial dataset with a MathML-to-LaTeX converter and a 4-stage dedup pipeline. • Fine-tuned Qwen 3.5 (27B) into a persona tutor on 85K SFT + 30K preference pairs, rubric 48% to 87%.

• Built a full-stack C2 controller for a red-teaming framework (gRPC + mTLS + Protobuf, 40% smaller payloads), JWT + RBAC for 50+ operators. • Only intern offered a remote continuation, with a 28% stipend hike. • Built an agentic, MCP-compatible red-teaming system on Qwen 2.5 (32B), LoRA + RLHF, with grounding and episodic memory. • Shipped an ML request filter blocking 95%+ of malicious traffic, plus a RAG CVE-intel system mapping host fingerprints to known vulns. • Deployed the C2 on AWS EC2 + MongoDB + Docker isolation behind Nginx, Cloudflare strict HTTPS, and DNS redirectors.

Attachments

Letter of Recommendation

Attachments

CyStar Completion Certificate.pdfAttachment

• Built core features for Samvadini, the Government of India's multilingual messaging platform. • Shipped an offline-first message queue (IndexedDB sync) cutting failed deliveries 85%; WebSockets + Redis pub/sub under 100ms. • Linked 20+ React Native screens to the backend, dropping page load 3s to 0.75s (75% faster). • Cut P95 API latency 47% (request tracing, N+1 elimination) and ran live MongoDB migrations on 100K+ docs with zero downtime.

Attachments

Samvadini Messenger

Extracurricular

02Selected work^[13]

You describe a concept in plain English and Stitcher writes the Manim code, runs it, and renders an actual animated video. Around 8K signups and 50+ people using it daily.
A crew of LLM agents handles codegen across 10+ providers and debugs its own mistakes in a retry loop. RAG over the Manim docs killed about 90% of the hallucinated code.
Every AI written script runs inside a locked down sandbox (bubblewrap + seccomp) so untrusted generated code can never touch the host machine.
Built a real time collaborative timeline editor with Yjs CRDTs over WebSockets. Basically Google Docs, but for editing video.
Payments are race safe: the Razorpay verify call and the webhook can both fire and you still get charged exactly once, thanks to an atomic Mongo update.

Attachments

Demo video↗

try it live!↗

TypeScriptPythonManimLangGraphLLaMa 3.1 8BRAGBullMQRedisMongoDBFFmpegRazorpay

Trained a 99M parameter transformer from scratch, same architecture family as Llama and Mistral (RoPE, grouped query attention, SwiGLU, RMSNorm).
Wrote my own byte level BPE tokenizer with a 32K vocab and built the full 4 billion token training corpus myself, deduping millions of near duplicate docs with a Bloom filter.
Full post training stack: supervised fine tuning, then DPO, with an LLM as a judge harness scoring relevance, persona and coherence on a held out set.
The whole thing runs client side in the browser (PyTorch to ONNX to WebAssembly, int8 quantized) at 100+ tokens/sec with nothing running on a server.
Taught it to use tools mid generation: it emits plain text calc tags the runtime catches and solves, so it does arithmetic cleanly and gets ~56% on harder word problems.

PyTorchTPUBPERoPEGQASwiGLUSFTDPOONNXtransformers.jsWASM

Built a 14 service agent system from scratch (gateway, main agent, 5 connectors, cron, a human in the loop dashboard) that runs 24/7 on a single machine.
Read the OpenClaw architecture and implemented its gateway myself: FIFO ticket dispatch over Redis, heartbeat based eviction, and recovery of in flight work if a worker dies mid task.
The core agent loads skills on demand, remembers past context with a vector memory (mem0), and can switch between LLM providers at runtime with automatic fallback when one fails.
Gave it real hands on the web with Skyvern for browser automation and Tavily for live research, so it can actually go do tasks instead of just talking about them.
Backed by a 381 test monorepo with real infrastructure integration tests spun up through Testcontainers, not mocks.

TypeScriptNode.jsBullMQRedisPostgresNext.jsPI Agentmem0SkyvernTavilyPM2

Attachments

Letter of Recommendation

Attachments

CyStar Completion Certificate.pdfAttachment

Attachments

Samvadini Messenger

ArjunVerma

How a single file upload gave me my college's entire server

ClashPilot: can an LLM actually play Clash of Clans?

Building Stitcher: an AI that animates anything you type

CHAD: building a 99M-param Hinglish texting bot from scratch

ArjunVerma

How a single file upload gave me my college's entire server

ClashPilot: can an LLM actually play Clash of Clans?

Building Stitcher: an AI that animates anything you type

CHAD: building a 99M-param Hinglish texting bot from scratch