Ankit Dahal

About Me

Hello! I am a software engineer with experience building and scaling applications for E-commerce and marketplace platforms serving both millions of users, and internal teams.

Recently I made the decision to step away from traditional software engineering roles to fully dedicate my time to learning and exploring artificial intelligence/machine learning, driven by a strong personal interest and a desire to tackle new technical challenges.

My goal is to transition into applied AI engineering, where I can focus on building with, training, and scaling deep learning models in real-world applications.

Work Experience

Resume available upon request

Senior Software Engineer

Brownells (Contractor) Jan 2022 - July 2025

    Product development - business process related

  • Designed and implemented reliable core communcation services between ERP and E-commerce for business critical tasks: orders, inventory levels, customer data, payments, etc.
  • Integrated across company-wide business functions such as warehouse/retail management systems, fraud, compliance, sales and promotions, and much more, working with dozens of internal users to understand and improve their processes.
  • Empowered non-technical teams by building internal tooling for technical self-remediation, orchestration, and reporting, improving productivity across teams.
  • Infrastructure responsibilities - scaling for millions of customers

  • Reliability - data consistency across distributed systems (customer data, handling orders/payments up to 9 figures a year)
  • Performance - caching with Redis to save on tens of millions of calls, full-text search on 100M application logs
  • Observability - telemetry, monitoring, alerting, traceability.
  • Operations - CI/CD pipelines, emailing strategy (SMTP), terraform, identity, cloud migration, application and database scaling

Software Engineer

Country Maid (Contractor) May 2021 - Jan 2022
  • Integrated tax services through TaxJar to meet marketplace sales tax compliance, allowing for sales & expansion across all 50 states.
  • Rehauled customer facing store, removing customer friction to checkout, adding visual aid, and integrating various digital wallets e.g. apple & google pay
  • Worked on an effort to maximize customer experience by clearing out 90% of customer facing bugs that were tracked over a 2 year period.

AI/ML Projects

Legal-based agentic RAG

Ideas include data sourcing, pre-processing, retrieval, reranking, finetuning & scaling training, optimization, evaluation, agents

Click here for project diagram

Click here for project notes

Code is available on request

(embeddings)bge-m3/gemini-embedding-001 (indexing)ChromaDB: HNSW tuning, Elasticsearch: BM25 (hybrid search)RRF, convex combination (reranking)bge-reranker-large (training)Pytorch, Docker, Google Kubernetes Engine, 560M parameters (agents)LLM: Gemini-2.5, conversational agent, search agent, planning, self-triage (evaluation)recall@k, precision@k, f1_score, mean_reciprocal_rank (MRR), mean_average_precision (MAP), agent evals

Current/future project ideas

Recently implemented the original "attention is all you need" paper from scratch. Click here to view

Currently working towards LLM training. Future plans are to reiterate my implementation with modern techniques seen in the LLaMA & Deepseek papers: rotary positional embeddings (RoPE), SwiGLU, multi-head latent attention, RMSNorm, etc.


Future interests include optimizations/scaling, mixture of experts architecture, computer vision and multi-modality.