Ankit Dahal

About

I took the decision to leave my role in 2025 to create a path for myself to pivot into AI/ML engineering. Years of foundational AI knowledge have been compressed into months of strenuous study and application through personal projects. This has allowed me to move exponentially faster than I would have been able to otherwise, building a strong understanding of modern AI from first principles.

I am looking for technically challenging problems that matter. I strive to be a strong generalist, diving deep when needed, and am passionate about a wide range of topics from AI, distributed systems, and the sciences.

Projects

Browser use agents & harnessing

AI Safety Feb 2026 – Current

Training browser use agents to complete web tasks and captcha solving

View details

Training browser use agents to complete web tasks and captcha solving, where modern models have a near 0% success rate on difficult captcha tasks.

5+ Platforms
AI Safety Domain

Technology: Qwen3.5-397B-A17B (multi-modal/VLM), playwright browser harness, sft & reinforcement learning, python

Adversarial AI Environment harnessing Platform analysis Agentic systems AI safety
Project Outline →

Llama 3.1 8B Instruction Tuning

Fine-tuning Jan 2026 – Feb 2026

LoRA fine-tuning with +52% IFEval improvement using ~$10 compute.

View details

LoRA fine-tuning of Llama 3.1 8B for instruction following. Achieved +52% IFEval improvement (200 → 305/834) with ~$10 of compute across 4-bit, 8-bit, and BF16 quantization experiments.

+52% IFEval gain
~$10 Compute
8B Base model
LoRA Unsloth Quantization IFEval W&B
Code →

Large Language Model Pretraining

Pretraining Dec 2025 – Jan 2026

450M parameter transformer inspired by the gpt and llama papers, pretrained on 10B tokens using 8xA100 GPUs.

View details

End-to-end pretraining of a custom, self-built 450M parameter GPT/Llama-style transformer on FineWeb-Edu (10B tokens). Trained on 8xA100 GPUs using Distributed Data Parallel with custom training loops and memory optimizations.

450M Parameters
10B Tokens
8x A100 GPUs
PyTorch, DDP Custom training loop Chinchilla scaling law RoPE, RMSNorm, SwiGLU, GQA/MHA, KV cache Flash attention
Code → Model Checkpoint →

Transformer from Scratch

Fundamentals Nov 2025 – Dec 2025

Clean PyTorch implementation with modern LLM improvements.

View details

Clean PyTorch implementation of "Attention Is All You Need," extended with modern LLM architectural improvements: RoPE, SwiGLU, RMSNorm.

PyTorch MHA SwiGLU/ReLU RMSNorm
Code →

Legal RAG System

RAG / Agents Jul 2025 – Oct 2025

Production-grade agentic RAG for legal documents with hybrid search (semantic + lexical).

View details

Production-grade agentic RAG system for legal documents. Full pipeline from data sourcing through retrieval, reranking, and agent orchestration. Hybrid search combining semantic embeddings (bge-m3) with lexical retrieval (BM25).

bge-m3 ChromaDB Elasticsearch bge-reranker Gemini-2.5 Docker GKE

Embeddings: bge-m3, gemini-embedding-001

Indexing: ChromaDB (HNSW), Elasticsearch (BM25)

Retrieval: Hybrid search with RRF, convex combination

Agents: Conversational + search agents, planning, self-triage

Training: 560M parameter model fine-tuning

Code → Training Infrastructure →

Writing

Technical Writeup

Building a Production-Grade Legal RAG System

End-to-end design notes covering hybrid retrieval architecture, embedding strategies, reranker integration, agentic orchestration, and lessons learned building a RAG pipeline for legal documents.

Open to Opportunities

Looking to join teams building interesting AI systems.

View Resume GitHub LinkedIn