Curriculum Vitae

Mohammed Akram Khan Lodi · Chennai, India
[email protected] · akramlodi.com · GitHub · LinkedIn

Research interests

Recursive self-improvement (RSI) · Agentic systems & harness engineering · Large language models (LLMs) & recursive language models (RLMs) · Long-horizon reasoning · Efficient AI systems

Education

B.S. Abdur Rahman Crescent Institute of Science and Technology, Chennai, India
B.Tech in Computer Science and Engineering · Jun 2024 – Jun 2027 (expected)
Transferred from the University of Victoria

• GPA 9.77 / 10 · Rank 1 in batch
• Coursework: Analysis of Algorithms; Data Structures; Theory of Computation; Artificial Intelligence Techniques; Natural Language Processing; Operating Systems; Network Security and Cryptography; Discrete Mathematics

University of Victoria, Victoria, British Columbia, Canada
B.Eng in Computer Engineering (first year) · Sep 2023 – Apr 2024

• GPA 7.89 / 9
• Coursework: Fundamentals of Programming with Engineering Applications; Matrix Algebra for Engineers; Calculus I; Calculus II

Publications & papers under review

• Self-Harnessing Recursive Language Models
  William Stanford, Mohammed Akram Khan Lodi, Jaloliddin Boymakhammadov, Eliaz Calvar, Simon Coumes
  Submitted to the First Workshop on Meta Agents: Managing Agents that Manage Agents, NeurIPS 2026 (under review)

Presentations

• The Carbon Cost of Reasoning: Benchmarking Energy Efficiency in Fine-Tuning Gemma 4 for Mathematical Logic
  National AI Summit on Industry 5.0, 2026 · B.S. Abdur Rahman Crescent Institute of Science and Technology, Chennai
  Paper ID AIS 052 · Accepted & presented, Apr 2026 · Proceedings

Research experience

Researcher, Self-Harnessing Recursive Language Models · Jul 2026 – present
Algoverse AI Research Program · Advisor: Dr. Simon Coumes

• Investigating automatic optimization of recursive language model (RLM) harnesses and their generalization across tasks and context lengths.
• Designed and evaluated RLM experiments across long-context benchmark suites, independently identifying and testing datasets suited to evaluating recursive decomposition, sub-call behavior, and harness-level failure modes, including OOLONG-Pairs, OBLIQ-Bench, and related datasets, using open-weight models.
• Contributed to the Self-Harness optimization pipeline, implementing and testing the harness-proposal workflow that converts recurring failure patterns into targeted modifications to RLM orchestration policies.
• Debugged and validated the experimental and evaluation infrastructure across benchmark configurations, supporting the full loop of weakness mining, harness proposal, proposal validation, and optimized-harness evaluation.
• Analyzed optimized harnesses against initial and manually engineered baselines on correctness, recursive behavior, generalization across task configurations, and computational efficiency.

Independent Researcher, The Carbon Cost of Reasoning · Apr 2026 – present
B.S. Abdur Rahman Crescent Institute of Science and Technology

• Designed a controlled comparison of Full Fine-Tuning, LoRA, and QLoRA on Google’s Gemma 4 E2B-it (~5.1B parameters) on GSM8K, measuring mathematical reasoning accuracy alongside environmental cost.
• Implemented energy and emissions tracking with CodeCarbon, recording GPU, CPU, and RAM energy (kWh) and estimated CO₂e per training run from the cloud region’s grid carbon intensity.
• Designed reproducible experiments with fixed prompt templates, dataset splits, and evaluation protocols: Full FT, LoRA (ranks 8/16/32), and QLoRA (4-bit NF4; ranks 8/16/32) across core and ablation runs on AWS EC2 (NVIDIA A10G and L40S).
• Introduced the Green Gap, the point where marginal reasoning improvement per additional kWh declines sharply, and the Reasoning Efficiency Index (REI), accuracy gained over zero-shot per kWh consumed.

Industry experience

Co-Founder, Airbil (AI-native agency) · Dec 2025 – present

• Built multi-agent AI systems using LLMs, FastAPI, and PostgreSQL/Supabase for enterprise procurement, sales, and customer-engagement workflows.
• Developed agent evaluation and monitoring infrastructure to measure task performance, tool usage, failure modes, latency, and reliability.
• Engineered scalable agent orchestration and API integrations connecting ERP/CRM and messaging platforms into unified agentic workflows.

Software Engineer, Skillsync (AI-powered hiring platform) · May 2025 – Jul 2025

• Developed an AI interviewer that conducts candidate interviews and generates automated candidate analysis and evaluation.
• Built AI-driven candidate assessment and an ATS for automated resume screening, candidate ranking, and structured evaluation.
• Designed a concurrent, worker-pool task scheduler in Node.js with fault-tolerant job queuing, processing 10,000+ asynchronous tasks per month with zero downtime.
• Re-architected the PostgreSQL schema and moved data access to normalized REST endpoints, cutting query latency by 50% under production load.

Co-Founder, PocketLink (AI-powered link-in-bio platform) · Feb 2025 – Dec 2025

• Architected cloud infrastructure on Docker, Kubernetes, and Azure serving 20,000+ users.
• Designed distributed inventory and payment-processing pipelines handling concurrent checkouts with real-time data consistency.
• Deployed an LLM-based conversational system using Gemini and LangChain with per-user context orchestration.

Projects

NDA Analyzer: Multi-Agent RAG System · Apr 2026 · code

• Four-agent LLM pipeline for employment-NDA analysis: clause extraction, loophole detection, legal-compliance verification, and fix generation.
• RAG over Indian legal statutes, NDA templates, and HR policies using ChromaDB and sentence-transformer embeddings.
• Full-stack app in Next.js and TypeScript with Gemini 2.5 Flash Lite, streaming responses over Server-Sent Events.

Technical skills

• Programming: Python, TypeScript/JavaScript, C, SQL
• AI/ML: PyTorch, LangChain, RAG, embeddings, fine-tuning, LoRA, QLoRA, LLM APIs
• AI systems: AI agents, multi-agent systems, recursive language models, LLM harnesses
• Web & backend: Next.js, React, Node.js, FastAPI, PostgreSQL, Supabase, REST APIs
• Cloud & infrastructure: AWS, Azure, Docker, Kubernetes, GitHub Actions, microservices