Build and debug production RL environments for training and evaluating agents across heterogeneous software tasks.
Own task design, state transitions, tool execution, and reward/verifier logic for environments used in model training and evaluation.
Diagnose whether a failure came from the model, the environment, the evaluator, or the execution layer — then turn the discovered shortcut or regression into durable test coverage.
Validate environment quality and reproducibility so benchmarks expose genuine capability gaps rather than superficial pass/fail behaviour.
Independent research lab. Publishes papers, open weights and research code across interpretability, reasoning and decision models.
Published three research artifacts with DOIs: a GPT-2 interpretability study, LEMMA (neuro-symbolic maths), and Noma (an open decision model).
Shipped Noma with open weights on Hugging Face under MPL-2.0, a published PyPI package, and ~16 ms end-to-end latency on a single GPU.
Maintains LEMMA, a Rust neuro-symbolic mathematics platform with 572 transformation rules and a released policy checkpoint.
Research
Open weights
Interpretability
Rust
Product & AI Systems (Contract)
current
Chalkbyte / Pyndra · 2026 – Present · Remote
Turn ambiguous business and user needs into scoped engineering work, then build the AI capability against it.
Research user and business problems, decide what should be built, and translate partially defined needs into technical scope rather than waiting for a complete specification.
Iterate AI capabilities with engineering from real stakeholder feedback — applied LLM workflows, experimentation, evaluation and product integration.
Led a pod on RL workflows, agent-evaluation frameworks, task quality and data curation for frontier-model development.
Designed and reviewed evaluation harnesses and failure cases, focused on whether tasks and graders measured the intended capability rather than an exploitable shortcut.
Coordinated review and delivery across contributors, and contributed upstream to Bespoke Curator with normalized Mistral batch finish_reason handling plus regression coverage.
RL
Agent evaluation
Data curation
Team lead
Researcher — Vision-Language-Action Robotics
IKSHMA Lab, IIT Mandi · Mar 2026 – Aug 2026 · IIT Mandi
Embodied and vision-language-action robotics research: multimodal perception, robot-learning workflows, simulation and model evaluation.
Explored reproducible VLA experimentation and robotics data pipelines across language, visual observations, actions and simulated environments.
Work in this area produced GS-DroneGym, a published drone-learning environment with photorealistic Gaussian-Splatting rendering.
Rust ML infrastructure across training and inference pipelines, boosting components, evaluation harnesses and model packaging.
Build reproducible evaluation harnesses, dataset/versioning conventions and production-oriented model packaging.
Develop boosting/GBDT components with emphasis on low-latency inference, memory efficiency, drift-aware monitoring and deterministic rollback paths.
Rust
GBDT
ML infrastructure
Reproducibility
AI Mentor & Technical Lead
Edge Lab · Jun 2025 – Sep 2025 · Remote
Led code review, testing practice and technical mentorship.
Helped teams ship production-ready projects with clean interfaces and deployable pipelines.
Mentorship
Code review
Founder & Lead Developer
current
UPAI Platform · May 2024 – Present · Remote
LLM orchestration platform spanning 100+ models with unified APIs and predictable contracts.
Built secure backend services with rate limiting, async jobs, retries, caching and PostgreSQL-backed state.
Built an LLM orchestrator and a benchmarking leaderboard for NVIDIA VIM.
LLM orchestration
Backend
PostgreSQL
Co-Founder & CTO
Boostra AI by GradeBoostHub · Feb 2020 – May 2026 · Remote / Hybrid
Owned product and engineering end-to-end for a production EdTech platform.
Database design, APIs, authentication, dashboards, production deployments, observability and incident response.
Operated Docker-based infrastructure with structured logging, monitoring, backups, RBAC, JWT/OAuth and hardened service boundaries.
CTO
Backend
Infrastructure
EdTech
//Speaking
Accepted40-minute technical talk
From Paper to Working System: Reproducing AI Research When the Weights Aren't There
Opportunity Open Source Conference 4.0 · IIIT Allahabad
What it actually takes to get a paper running when the checkpoint was never released — reading a method closely enough to rebuild it, and knowing when your reproduction has diverged from the claim.