Skip to content
View vnktadithya's full-sized avatar
🎯
Focusing
🎯
Focusing

Highlights

  • Pro

Block or report vnktadithya

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
vnktadithya/README.md

Venkat Adithya

I build LLM agents and the evals that show whether they work. B.Tech AI & Data Science @ Shiv Nadar University · 2027


Harness-level bug fixes in the OpenHands agent SDK, each shipped with regression tests.

  • #4435 · merged — The context condenser's max_tokens wasn't inheriting the LLM's effective max input tokens, silently truncating context on specific launch paths.
  • #4566 · merged — The global load_memory preference reached only one of three conversation launch paths. Centralized the fix; 5 regression tests.
  • #4876 · merged — load_skills_from_agent_server() discarded the caller's AgentContext. A follow-up bug I found and filed after #4566; 5 regression tests.
  • #5690 · in review — Concurrent plugin requests could corrupt .installed.json. File lock plus atomic write.

Projects

RCA LoRA Scaling Study — does more fine-tuning data help? Controlled QLoRA ablation on Mistral-7B: training set scaled 20 → 474 examples, with LoRA config, quantization, and prompt format held fixed. Scored by a 3-tier harness I built: JSON validity, embedding similarity, calibrated LLM-as-judge. The first two tiers called it a near-tie. The judge scored the 474-example model lower (2.00 vs 2.21, n = 53 held-out), and manual cross-checks confirmed fabricated technical content in 8 of 9 cases. More data didn't help at fixed adapter capacity.

ReAct Agent — a raw-Python ReAct loop on a local Llama-3.2-1B-Instruct GGUF via llama-cpp-python, CPU-only. No agent framework: hand-rolled action parser, tool dispatcher, step budget, and a circuit breaker that halts on the first repeated (action, param) pair anywhere in run history.

KV Cache VRAM Calculator — the KV cache formula derived from first principles and verified on a T4 against Llama-3.2-1B-Instruct. It matched direct cache-tensor inspection exactly (25.31 MB), while torch.cuda.max_memory_allocated() reported 48.10 MB. Covers the GQA trap: using num_attention_heads instead of num_key_value_heads overestimates the cache 4×.

Siamese Bi-Encoder — a toy Sentence-BERT setup trained from an empty embedding table. Perfect test AUC across 3 seeds hid a test MSE 750–1,500× the train MSE.


Pinned Loading

  1. Repro-Agent Repro-Agent Public

    Python

  2. rca-lora-scaling-study rca-lora-scaling-study Public

    QLoRA fine-tune of Mistral-7B for root-cause-analysis extraction — an ablation testing whether more training data (20→474 examples) actually improves quality, scored across 3 independent metrics.

    Python 1

  3. ReAct_Agent ReAct_Agent Public

    Python

  4. LLM-inference-math LLM-inference-math Public

    Jupyter Notebook

  5. siamese-bi-encoder-from-scratch siamese-bi-encoder-from-scratch Public

    Sentence-BERT's siamese bi-encoder rebuilt from scratch in PyTorch on root-cause descriptions of real bugs: BoW baseline, 3-seed AUC/MSE evaluation, and an OOV probe where unseen words push unrelat…

    Python

  6. Learning-Log Learning-Log Public

    Python