Headline

Artificial Intelligence Engineer @ Bitqcode Quantitative Capital |

About

A graduate of Birla Institute of Technology, Mesra with a Bachelor of Technology in Electrical, Electronics, and Communications Engineering. Currently contributing as an Artificial Intelligence Engineer at Bitqcode Quantitative Capital, focusing on advanced AI workflows and knowledge graph research. Core competencies include predictive modeling, natural language processing, with demonstrated expertise in building scalable data ingestion platforms and retrieval-augmented generation (RAG) systems. At Bitqcode, collaborated with a team to develop a self-improving multi-agent LangGraph pipeline and a comprehensive RAG-based research layer, enabling efficient knowledge extraction and strategy validation. Passionate about leveraging technology to drive innovation, while fostering collaboration and creating impactful solutions. Dedicated to integrating advanced analytics with organizational goals to deliver meaningful outcomes.

Location

Bengaluru

AI stack

claude code

hermes agent

langgraph

langsmith

notebooklm

opencode

Work experience

Artificial Intelligence Engineer at Bitqcode Quantitative Capital, Jul 2026 - Present · 4 mos, Bengaluru, Karnataka, India, Built, alongside a team of AI engineers, a self-improving multi-agent LangGraph pipeline (Claude via AWS Bedrock) running 24/7 at 30-min cadence with bounded retry, convergence gating, and crash-safe stateless processes; persisted 96+ strategies with full trade logs and provenance to PostgreSQL, producing the platform’s first validated directional strategy. Spearheaded a self-proposed RAG and knowledge-graph research layer over ∼8,000 papers (arXiv/NBER), architecting schema-bounded claim extraction into pgvector (HNSW) and Neo4j (∼50K nodes, ∼500K edges) serving 4 retrieval use cases with deterministic full-rebuild projection. Designed a unified data-ingestion platform (probe → fetch → validate → stamp → upsert) with ingest-twice determinism across 8+ sources – including a 500K+-event economic calendar and 10+ years of 1-minute FX tick and options pricing data – guarded by a staleness interlock blocking research on stale data. Proposed and led the build of a synthetic evaluation benchmark of 1,250 queries over 200 sampled papers with AI-generated ground-truth labels (10% human-verified, 0.85 annotator agreement). Evaluated the RAG pipeline against the benchmark, scoring recall@5 0.91, MRR 0.87, and NDCG@10 0.89, with 96.4% per-node output correctness across the agentic pipeline.

Data Scientist at Tredence Inc., Sep 2024 - Jun 2026 · 1 yr 10 mos, Gurugram, Haryana, India, Engineered, in coordination with Thorne stakeholders, a churn prediction pipeline using logistic regression for 136K customers; achieved 78% accuracy and 0.87 precision with rankings written to Redshift for CRM activation. Architected, alongside fellow Tredence data scientists, a two-stage CLTV system for 3, 6, and 12-month horizons; improved spend R2 from 0.54 to 0.59 and precision up to 0.90. Developed the end-to-end RAG embedding pipeline for Taia on AWS Bedrock, indexing a 1,500-product catalog into OpenSearch for sub-second semantic retrieval. Designed a middleware API gateway with 11 FastAPI endpoints and Valkey caching, maintaining a 3.1-second average end-to-end query time. Implemented an async batch pipeline on AWS Batch using GPT-4.1 to evaluate 12,000+ daily chatbot messages across accuracy, relevance, and tone dimensions.

Analyst at Sharaf DG, Jan 2024 - Jun 2024 · 6 mos, Developed a Selenium validation engine to audit supplier catalogs, reducing manual screening time by approximately 80%.

Education

Bachelor of Technology, Electrical, Electronics and Communications Engineering, Birla Institute of Technology, Mesra

Public links

github: https://github.com/Akshat-2011